test(bc-e2eh, mp-yqxn.2): e2e harness and a clumsy-operator review of knockout and mixed by gitrgoliveira · Pull Request #442 · gitrgoliveira/bracket-creator

gitrgoliveira · 2026-09-24T08:19:55Z

Summary

Ships the local e2e harness (scripts/e2e/) and the persona-audited operator journeys for knockout and mixed (pools + knockout) competitions, individual and team, plus two shared-shiaijo journeys. make e2e builds the binary, starts mobile-app on a free port with a fresh data folder for each spec file, and drives the real UI in Chromium on an iPad Air landscape (1180x820, (pointer: coarse) asserted) and a spectator phone (390x844). Every competition is seeded through the UI: the first-run form, the login form, the create wizard, the roster paste box, Generate draw and Start. Nothing is added to make go/test or to GitHub CI.
Ships three app fixes, bc-ltap, bc-sdtp and bc-xrpl, all found by this PR's own runs and fixed here on the operator's word, and one operator-ruled change, bc-cmpw (one admin password), which removes the separate destructive-ops password and sets self-run rules (section "The fourth change"). Everything else under web-mobile/, internal/, cmd/ and docs/ is unchanged, apart from the docs and spec lines that describe these rules. The review's output is beads parented to mp-yqxn.2. Each functional defect a journey hits is a test.fixme('<bead>: ...') in its spec, asserting the fixed behaviour, so the bead's own PR turns it into a plain test.
The knockout and mixed review is fully fixed on main. fix(bc-tmfn): operator fixes: team bouts, withdrawals, knockout rounds and the Bracket page #444, fix(bc-kfup,bc-htcr,bc-sbq,bc-wnam,bc-kclr,bc-strt,bc-kenu,bc-plcl): P1 operator fixes, a queued match keeps its score, and the 2026-09-26 rulings #447, fix(bc-dtfn, bc-dtip, bc-fsnp, bc-emsl, bc-sync): a bounced tap, a fusensho or a reload no longer changes a recorded result #450, fix(bc-dhas, bc-k404, bc-xlcl, bc-kdsc): representative bout vs autosave, self-run lineups, White-left Excel, Kachinuki Detail #451, fix(bc-rfsw,bc-offl,bc-rboq,bc-flst,bc-kheb,bc-mrgc,bc-hlck): merge match writes by change, held results visible, correction refusals, kachinuki encho #453, fix: iPad operator fixes; a team's lineup carries between matches #458, fix(bc-dsid, bc-otpl, bc-htsd, bc-dlck, bc-kosc, bc-crpn): operator review fixes from PR #442 #462, fix: five operator-review bugs: auto-advance refusal, court-call double tap, daihyosen representative, team winner cue, drawn recent results #463 and feat: kachinuki advisory line and recorded bouts, 3rd-place labels, form labels, and the open #442 review beads #467 closed 51 of its 53 beads (50 fixed, bc-kpnl superseded by the 2026-09-26 "do not cascade" ruling, its two tests deleted). Every one of those fixmes is now a plain test that guards the fix, and the journeys follow fix(bc-dsid, bc-otpl, bc-htsd, bc-dlck, bc-kosc, bc-crpn): operator review fixes from PR #442 #462's deliberate changes (Correct is refused at the console while a bout is live, bc-crpn; the withdrawal prompt picks no side, bc-dsid; the hantei buttons pick and Finish records, bc-htsd).
New since the review: the shared-shiaijo journeys (bc-shcr, bc-mvsj). shared-shiaijo.spec.mjs runs two competitions on one court (S1 to S9, P1 to P8) and moved-match.spec.mjs moves a waiting match to another court, watched on two devices. Their findings went to the operator with screenshots; the eight confirmed beads are filed under mp-yqxn.2 and each has a test.fixme in these two specs (section below).
Harness review (2026-10-11): one server per spec file; a console policy (a console error or uncaught page error fails the test unless it is on a short, reasoned allowlist; warnings are recorded); selectors by label and role rather than layout; persona helpers that pin the element they wait on; the coarse pointer restored in place after a fullPage shot. Details under Harness notes.
Walked as a distracted, clumsy operator. For every operator action each applicable variant was performed in the browser: V1 neighbour tap, V2 double tap (and a re-tap while "Saving…"), V3 interruption (reload, back, tab hidden, offline), V4 hasty dialog (tap the loudest button), V5 lost place. fixtures/clumsy.mjs enacts them, so the audit re-runs on every later PR.
Audit tables are PR comments, not in this body (GitHub's body limit). The three 2026-09-24 comments are the review as performed, with the reviewers' judgement; the comments posted on 2026-10-12 are the raw rows from a current run of all five journey groups.

Which PR closes which bead

This PR (test(bc-e2eh, mp-yqxn.2): e2e harness and a clumsy-operator review of knockout and mixed #442) closes eight: bc-e2eh (the harness), mp-yqxn.2 (the review), bc-shcr (the shared-shiaijo journey), bc-mvsj (the moved-match journey), bc-ltap, bc-sdtp and bc-xrpl (the three fixes below), and bc-cmpw (one admin password). They stay in progress until the merge.
Follow-up: bc-e2rv re-plans and re-verifies the harness against main once this PR has merged.
Closed already on main: the other 51 review beads, each by the merged PR marked against it in the findings table.
Open, each with a fixme here, closed by its own fix PR: bc-mvrq, bc-sypl, bc-stpf, bc-adna, bc-pubq, bc-pace, bc-seqc, bc-rnrm.

The fix: a tapped-back mark was lost (bc-ltap)

What happened. On the individual score editor, an operator who took a mark off and tapped the same mark back while the removal's autosave was still out lost the tap: the board showed it, then the removal's echo from the server wiped it, and the tap's own save named no change. The reverse (add, then remove quickly) brought the mark back. Found when J1 failed once in a fresh clone; reproduced 6 of 6 with the removal's request held 150 ms.

Root cause. The editor judged unsaved work by comparing the board with the server's current value (pointsEdited, isDirty). A tap that puts back the value the server still shows (its echo of the removal not yet arrived) reads as untouched, so useAdoptFromServer took the echo over the newer tap.

Fix. The editor records when the operator last edited the board (points, fouls, overtime, draw) in the server's clock frame and never adopts a server copy stamped before it (adoptServerScore, admin_scoring_individual.jsx), the rule the team sheet's rows already follow (rowTakenFromServer, bc-kclr). Taking the server's copy clears the stamp, so a change made elsewhere after the last tap is still followed. The team sheet was checked and is not affected.

The second fix: a double tap on Start match started the next match (bc-sdtp)

What happened. On the court console, a double tap on the Up next card's Start match could start the wrong match. The first tap started semi-final 1; on a fast link that start landed inside the app's 400 ms bounce window, the card re-rendered to show semi-final 2 with its Start match enabled, and the second tap (289 ms after the first) sent semi-final 1 back to the queue and started semi-final 2. Found by the shared-shiaijo journey (S3) in a --workers=1 run; the trace shows the two taps and the two start writes.

Root cause. The console's Start match buttons (the Up next card and every queue row) had no bounce guard: each was disabled only while its own match was starting, so the bounce landed on the next match's enabled button. Call to court on the same card was already guarded (bc-cdbl).

Fix. Every Start match on the console shares one keyless bounce guard (startTapRef through acceptTap, tap_guard.jsx, the one owner), so a bounce is ignored whichever Start it lands on; a deliberate tap after the window and a keyboard activation still start. start_match_double_tap.render.test.jsx fails without the fix (the bounce started the next match, from the card or from a queue row) and passes with it.

The third fix: a finish whose answer was lost was refused on resend (bc-xrpl)

What happened. When the server stored a Finish but the device never processed the answer (wifi dropping between request and reply, the iPad sleeping or the tab reloading right after the tap, a reply slower than the app's 12 s wait), the offline queue kept the write and sent it again with the same stamp, as designed. The server read the identical copy as a correction of a finished match with no reason and answered 400, so the operator was told a recorded result was rejected. Found by J6 in a fresh-clone run (an offline reload let the queued Finish out, and the page was gone before the answer arrived).

Root cause. The correction-reason rule treated an equal stamp over a finished match as a correction (applyCorrectionReason), while the queue relied on an equal stamp re-applying. Decisions already answered an exact resend as recorded (exactDecisionReplay); score finishes did not.

Fix (operator decision 2026-10-12, option A). An exact resend of a completed score write (same stamp as the stored result, and nothing it names differing from what is stored, read the way the write path would land it) is answered as recorded, with the stored match and no correction reason (engine.ExactScoreReplay, checked in applyCorrectionReason and answered in RecordMatchResultWithIneligibilityTx before any refusal; nothing is written to the match file, and the history gains a line with every group unchanged, as for a decision). A same-stamp write with a different result still needs a reason. The client also takes a settled queue entry out of browser storage at once, not at the end of the pass, which narrows the window. Limits: a team or kachinuki finish whose bout rows the server rewrote on the first landing can still read as different and keep the old answer; a self-run participant's resend still meets result_finalized first.

The fourth change (bc-cmpw): one admin password

Operator rulings 2026-10-12 and 2026-10-13. Commits dfb552d (the change), 3b45f33 (merge of main, #472), ac60a87 (round-3 fixes, the reset rule), and the reset Host check made after.

One password. The separate destructive-ops password is removed, and a signed-in organiser is never asked for the password again, except for Delete competition: it asks to enter the password again (promptDialog), the client sends X-Confirm-Password, the server checks it with the sign-in verifier (confirm_password.go), and a wrong one is 401 "Wrong password. The competition was not deleted."
Upgrade note. TOURNAMENT_ADMIN_PASSWORD_HASH and a stored admin_password are ignored; remove them.
Self-run. Completing needs no password; anyone can add and edit participants; the branding logo and league tie-break routes are organiser-only. Self-run has no check-in: Competition.CheckInApplies is the one owner the draw, the Swiss field and the estimate ask, so the draw includes everyone, the setting and the registration desk are hidden, and the three check-in writes answer 409 check_in_not_in_self_run.
Completion. A competition with no knockout rounds cannot be completed while a match is unfinished (PoolMatchesSettled), nor a Swiss one while rounds remain (SwissRoundsToDraw).
Reset. In file mode POST /api/tournament/reset answers only a request made on the server machine: a loopback peer (never judged on forwarded headers) and a loopback Host (closes DNS rebinding). Otherwise 403 reset_local_only, which the /reset page shows. Locked mode answers 404.
Review fixes. The exact-replay probe (bc-xrpl) runs the merge on a copy, so held groups are excluded, and covers kachinuki, engi and team shapes. A participant's exact resend of a landed finish is answered as recorded with no history line. Both editors clear an edit stamp once a later save lands (landedAutosave). The file-mode password compare is constant-time.
Known gap. Swiss has no Complete competition button, so no podium; recorded on mp-yqxn.4 / bc-e2rv.

Shared shiaijo and the moved match (bc-shcr, bc-mvsj)

Two competitions, Shared Individual (SI, mixed) and Shared Team (ST, knockout), share Shiaijo A; SI also has B. Operator rulings on the findings (2026-10-11 and 2026-10-12), with the screenshots they were made from:

Bead	Ruling	Fixme
bc-mvrq	Moving a running match to another shiaijo warns first; once confirmed it stops running, keeps its score, and goes first in that court's queue	`shared-shiaijo.spec.mjs`, last test
bc-sypl	The score editors drop the "Synced" pill: the header's Connected pill and its held-writes count are the one signal	`shared-shiaijo.spec.mjs`
bc-stpf	Option A: a Start is checked before the running bout is sent back	`moved-match.spec.mjs`
bc-adna	Admin views do not start or advance matches and are not blocked by court status; the court console is where scoring and its queue live	`moved-match.spec.mjs`
bc-pubq	Public screens follow the court operator's queue on a shared shiaijo	`shared-shiaijo.spec.mjs`
bc-pace	The schedule adapts when courts run late or ahead	`shared-shiaijo.spec.mjs`
bc-seqc	Competitions sharing a shiaijo run one after the other by default, start times adjusted as the day goes	`shared-shiaijo.spec.mjs`, first test
bc-rnrm	"Run now" on a Later row is the wrong button: remove it	`shared-shiaijo.spec.mjs`

Not filed, by ruling: the court-move control on a queued match is fine as it is (a queued match can always change court), and the amber "Switch to" line already proposes the next match by schedule.

Walkthrough: the five-step operator path

Update, 2026-10-11. Main has since merged #447, #450, #451, #453, #458, #462, #463 and #467 as well, closing 42 more beads (marked in the table below). Where the walkthrough describes one of those beads, it describes the build as reviewed, not main.

Update, 2026-09-25, after main merged #444. The walkthrough below is the review as performed (at 548a936). These statements in it no longer hold on main:

§1 and §4: "Copy from previous match" is now at the top of each side (bc-cpdc). The remaining-matches panel's Award button records a fusensho (bc-wnam). The withdrawn competitor's next bout no longer offers Start match. Its queue row reads " withdrew: record the fusensho." with Record fusensho for , which records it and closes that bout (J2 now takes this route). bc-kpnl and bc-kfup still reproduce as filed.
§2: the shortcut hint is hidden on the iPad (bc-kbhn). A double tap on a fought kachinuki bout opens that bout (bc-kbrw). Ending a reopened encounter asks for no reason (door 2, and §5's "End match with a reason").
§3: closing a running match flushes the autosave with no discard prompt (bc-dscn). Untouched team bouts are no longer written as draws (bc-unfx). Finish refuses a team match while a bout has no result (bc-tmfn).
§4: refusals name the match instead of raw codes (bc-rawm). A tied individual knockout bout's Finish is disabled and reads "Needs a winner".

Fixtures, all through the wizard: individual mixed, 12 entrants, two shiaijo; individual knockout-only with the bronze match (joint 3rd unticked) and one on the default joint 3rd; team match, size 3, mixed; team match, size 5, knockout-only (daihyosen is knockout-only); kachinuki, size 5, knockout, one encounter driven eight bouts deep from /admin/shiaijo/:court. Per-row evidence is in the audit comments below.

1. Lineup entry

Individual: not applicable.
Team, size 3 and 5: size 3 shows bare positions 1, 2, 3 and size 5 shows Senpo, Jiho, Chuken, Fukusho, Taisho (positionsForSize), on all three surfaces. What held: a wrong fighter picked (V1) is repaired in two taps, and each list leaves out members placed elsewhere. A name typed on the score sheet renamed the blank member slot rather than minting a new member, and a re-tap on Save while "Saving…" wrote once. What did not: a double tap on the Lineups page's Add dismisses the confirm it just opened (bc-cfbd). A reload before Save loses the picks with no prompt (bc-lnul). The panel's duplicate refusal names the wrong position ("Dai is already at 2." for Dai at 1) in one of three wordings (bc-lprf). The sheet's refusal renders below the IV/PW band, off screen (bc-lnrf).
Kachinuki: all 20 team members were named by typing into the court console's "Enter lineup" panel. Typed names are lost on reload (bc-lnul). "Copy from previous match" is at the bottom of each side, while the docs say the top (bc-cpdc). On a later bout's fighter list, a double tap picks the fighter and then scores the ippon underneath, and a focused name box will not reopen its list (bc-flst).

Ruling: three lineup entry points is intentional in design and defective in execution.

They are three layers, not three copies. The Lineups page writes the team's ROUND default. The at-court panel ("Enter lineup" on Up next, "Lineup" on each Upcoming row) and the score sheet's name boxes write the per-MATCH override, which wins. Each serves a different moment: before the day, at the table before the encounter, and during the bout (the only one on screen while scoring). The docs describe this layering, so having three is not the defect.
They disagree with no operator error. On a pool match the panel showed every Shiro box empty under "Inheriting round default", while the Lineups page held Aoki/Baba/Chiba and the sheet showed them. The panel fetches the match's exact round index. The sheet, viewer and TV board ask with fallback=best (evidence J3-lineups/07-f3-panel-round-default.png).
The disagreement becomes data loss through a realistic slip. Tapping Shiro's Save on the empty-looking side writes an EMPTY match override, and the sheet's Shiro names vanish for that match while the Lineups page still shows them. Repair took three re-entries on the sheet.
The panel is the first lineup surface a hurried operator meets ("Enter lineup" sits beside "Start match"), and it is the one reading a different layer. The Lineups page is not on the court console at all. Nothing on the sheet says which layer a name came from.

Filed as bc-lpfb. Fix direction: the panel reads through the same resolver as the sheet, and a side whose values did not change is never written.

2. Navigating bouts and matches

Chained navigation never left the court. Finish + Start Next stayed on Shiaijo A across the pool to knockout boundary (last pool bout, then QF M1, QF M2, SF M5, all on A). On the Scores tab, Prev/Next walked seven editors and every one was on Shiaijo B. The same match and court came back after every V3 interruption: reload, tab hidden, back then forward, and 0.6s offline. Points were kept. V5: the court name, the eyebrow (competition, pool, match N of M) and both names and numbers stayed legible at arm's length.
Order is wrong, though. In a pools + knockout competition every knockout bout is scheduled at the day's start time, so Finish + Start Next on court B ran a resolved quarter-final between the remaining Pool D bouts (bc-kosc). The console's arrow keys and Esc do nothing, while the shortcut hint advertising them shows on the iPad (bc-kbhn). After Correct, Start match on Up next starts the next match, but the panel keeps the finished correction and the running match is nowhere on the page (bc-crpn).
Team match: bout rows are 205px tall, so about 2.4 bouts fit at 1180x820. The IV/PW band sits below bout 5, and the match identity scrolls away while scoring bouts 3 to 5 (bc-tmfd).
Kachinuki, kachinukiVisiblePositions verdict (context seven bouts in): enough context for "who fought whom and how", not for "where do we stand". Only the live slot is editable, but every fought bout renders above it as a read-only row: bout number, both fighters with member numbers, the marks, the centre (vs, X or (E)) and the winner. Nothing states fighters left per side or who is up next, so under exhaustion the operator counts by eye to know End match is due. That is ruled an advisory line (bc-kfnl). From about bout 6, Record bout and End match are below the fold (bc-tmfd). The header keeps "(E) OVERTIME x2" on every regulation bout after one encho bout (bc-kheb).
Kachinuki, returning to a mis-scored earlier bout: yes, through three doors that work.
1. Mid-encounter, tap the fought row. It opens in place with full controls and saves as you go, and a winner flip raises the documented warning. The only cue is a small ▶, and a double tap opens the bout ABOVE it (bc-kbrw).
2. On a finished encounter, Correct, then Reopen match, fix, and End match with a reason. The fix propagated (PW 8-3 to 7-3, winner and bracket side unchanged).
3. On a busy court, the panel names the running match and offers the atomic requeue.
"Remove this bout" works and is recoverable. The dangerous exit is the one not built for a bout log: Send back to queue under a reopened encounter wipes it (step 5, bc-sbq). Who stayed on was correct in every performed case: wins, hikiwake, a taisho tie, removal and re-record, and a correction after reopen.

3. Score entry and the running total

Individual: ippon buttons meet 44px. V1 hit K instead of M on the same side, and one tap on the mark clears it; "Tap a scored mark to clear it" shows while a mark exists, so the undo is discoverable. V2 recorded M M, a bout decided 2-0 on one intended point (bc-dtip). A reload inside the 300ms autosave debounce loses the point while the pill already reads Synced (bc-sync). From the Pools tab, the editor closes on Start match and after every autosaved point (bc-plcl), and "Discard unsaved scoring changes?" discards nothing because the autosave already landed (bc-dscn).
Team match: the IV/PW totals are right but off screen (bc-tmfd). After the first ippon every untouched numbered bout is written as a hikiwake. It shows X on the TV board and the viewer, and "✓ Tie" on the sheet after a reload; a Tie tap on it then turns the draw OFF (bc-unfx). A pool encounter finished with bout 3 never scored counts it as an individual draw for both teams (IT 2 on the Pools tab). Per the operator ruling, every bout of a team match is fought, so Finish must refuse (bc-tmfn). A stray tap on an EMPTY mark slot clears the bout's Tie or Fusensho (bc-emsl). Fusensho toggle then untoggle, in place, restored the prior score. After a reload the untoggle cannot restore it, because the fusensho had already erased the other side's points on the wire (bc-fsnp).
Kachinuki: a double tap on Tie toggles it on then off. A double tap on Encho doubles the count, and bout mode has no stepper (by ruling, not re-litigated). Record bout while offline leaves the bout live with no "Not sent yet" notice; it lands on reconnect (bc-rboq).

4. Special decisions

Kiken chain (closing the withdrawn competitor's remaining matches): BLOCKS, three defects, all three must land. After Kiken - Voluntary on the court console the Remaining matches panel shows for about a second, then the court moves on and it cannot be reopened (bc-kpnl, also on a team pool). Where the panel is reachable, its Award button throws wname is not defined and records nothing (bc-wnam). Fusenpai on the withdrawn competitor's next pool bout (Scores tab, the same call Award makes) is refused with the raw code already_ineligible (bc-kfup). Starting that bout, or scoring the opponent and finishing, is refused by design, with the raw text "kiken-voluntary at Pool A-0" (bc-rawm). So the pool never completes and the knockout never seeds from it.
Side picker: Kiken - Voluntary's neighbour (V1) is Kiken - Injury, and it only opens a prompt (Cancel recovers). The withdrawal and no-show prompt preselects Shiro, and a hasty Record (V4) withdrew the wrong competitor, and on the team journey the wrong team, with a voluntary kiken, which the docs call permanent (bc-dsid). The sanctioned recovery is to re-record the kiken on the other side through Correct (operator ruling; docs bead bc-kkdc). It took five taps plus the forced confirm below.
The 409 decision_locked confirm, judged under time pressure: not comprehensible. Correcting a kiken after a later bout has started shows "Please confirm ... A subsequent match for one of these competitors has already started ... Proceed anyway?". It names neither the match nor what proceeding does. In the individual journey, proceeding left the real withdrawer's running bout on the court, unfinishable (Finish refused, only Send back to queue, where it stays unstartable). In the team journey, it silently swapped the Shiro team in the running final. The paragraph break collapses. An operator at the table either proceeds blind or cancels and loses the correction (bc-dlck). The red "Proceed anyway" button itself is app convention for an irreversible override and was not filed.
Fusenpai: a double tap recorded one decision and no duplicate write. Auto-advance started the next bout by itself on a court with nothing stuck. Where the next bout was unstartable, the court sat idle behind a 2.7s toast (bc-aadv).
Encho: tapping the word "Overtime" opens the glossary instead of the encho counter, contrary to the docs (bc-otpl). The counter is unbounded as intended, and minus floors at x1. The max_encho_exceeded arm of the 409 no longer exists (mp-m4bn removed the cap), so it was not judged. The + and - steppers (30x44) and the Encho started checkbox (13x13) are under the 44px floor (bc-tp44).
Hantei, individual: BLOCKS. SHIRO wins and AKA wins sit 6px apart. A neighbour tap committed the opposite verdict at once and advanced the bracket (bc-htsd). Correct then cannot fix it: no reason prompt appears, the toast reads the raw "correctionReason: correcting a completed match result requires a non-empty correctionReason", and no other UI route exists (bc-htcr). Finish on a tied 1-1 individual knockout bout shows the raw "bracket match m-r1-0: cannot mark completed with no winner ..." over the Finish button (bc-rawm).
Daihyosen, team match size 5: a tied encounter disables Finish ("Needs a winner"). Add representative bout appends the row in place with (DH) in its centre. There is no way to pick each team's representative, although the docs and the in-app hint promise one (bc-dhrp). Found on re-running after fix(bc-tmfn): operator fixes: team bouts, withdrawals, knockout rounds and the Bracket page #444: tapped within the last Tie's 300ms autosave window, the row never appears, because that pending write lands after the daihyosen and erases it (bc-dhas). A hantei neighbour tap on the team sheet is recoverable in one tap before Finish. The closed middle held: Ht rode the winner's side and the centre stayed (DH). By operator ruling the bracket card does not carry that Ht, and the docs will be narrowed (bc-tmdc).

5. Confirm and advance

Two-tap guard: a double tap (15 to 30ms) on Finish, Finish + Start Next or End match arms and commits in one gesture (bc-dtfn). The armed End match label, verdict included, is dark text on a dark fill (bc-cfcl). A re-tap while "Saving…" was ignored, with exactly one completion.
Did Finish land? (V5): only the names change in the editor. A team result decided by daihyosen, kiken or fusenpai shows no winner on the court's Completed list ("IV 1–1 PW 1–1" reads like a draw) (bc-tmwn).
Send back to queue: its loud button is the destructive one. On the individual console, in one of two runs, its copy said "No score has been entered, so nothing will be lost" with a synced M on the board. Under a REOPENED kachinuki encounter it says "Any score entered on this bout will be discarded" and wipes the encounter's whole six-bout log (bc-sbq). End match with a reason is the non-destructive exit, and it is on screen.
Offline confirm and advance verdict (J6, knockout-only, shiaijo C, bronze match on): Finish on the last semi-final with the context offline. The local bracket advanced correctly: the final and the bronze moved into Up next (keep this, mp-y3nk). But the advance unmounted the editor that holds the "Not saved yet" banner, and the header still read "Connected", so nothing on the page said the result was queued (bc-offl). The same bout was not offered again, so there was no double entry. A reload while offline gives a blank page; that is a documented gap owned by mp-wm4e and not refiled. On reconnect the queued Finish landed and the server bracket agreed: the final carries the winner and the 3rd place match the loser. The bronze bout is labelled only "10:55 · Veterans Knockout" and is absent from Later until both semi-finals finish (bc-bzlb).
One result on every surface (J7): the TV board and the viewer followed the court. A court editor left open ADOPTED a hantei correction made at the head table without a tap. A drawn bout never reaches the viewer's Recent results (bc-rrdr). A double tap on Call to court posts two identical "Now calling" banners to the public viewer (bc-cdbl).
Pools and Bracket tabs (J8): after Finish on the Bracket tab the operator lands on the read-only result card, one tap from the next semi. Edit result is 44px tall under a coarse pointer.

What passed

Chained navigation never left the court. The same court and match came back after every reload, tab switch, back/forward and short offline. Double taps on Start match, Record bout, Save lineup, Reopen match, Remove this bout and the collapse caret each acted once. The tap-the-mark clear is discoverable. Fusensho untoggle restored the score in place. Kachinuki's who-stays-on was right in every case. A kachinuki correction through Reopen propagated to the result and the bracket. Editors adopted a verdict recorded on another device. The queued offline Finish landed on reconnect.

Operator rulings made during this review (2026-09-24)

The harness lives in scripts/e2e/, reusing scripts/screenshots/lib, not in web-mobile/e2e. Seeding is UI-only, through the wizard every time. One PR closes bc-e2eh and mp-yqxn.2.
Every bout of a team match is fought, so there are no unfinished team matches. There are no "fixed-order" teams: the team sets its lineup before each match, so the formats are "team match" and "kachinuki". Filed as bc-tmfn.
Re-recording a voluntary kiken on the other side through Correct is the sanctioned recovery for a kiken recorded against the wrong side. Docs bead bc-kkdc.
The bracket card does not carry Ht for a team daihyosen hantei, and the docs are narrowed. Filed as bc-tmdc.
Kachinuki: one bout's encho does not mark the encounter (E), in the editor header or on the bracket card. bc-kheb was widened.
Kachinuki bout mode gets an advisory "fighters left / up next" line that never ends, arms or blocks anything. Filed as bc-kfnl.

Findings (severity-ranked)

Every child bead of mp-yqxn.2. Severity: BLOCKS a running tournament, SLOWS the operator, COSMETIC. Reviewers: A individual, B team match, C kachinuki; "smoke" and "harness" are the harness build. Each bead cites current file:line and its evidence screenshots. "(closed, #NNN)" marks a bead closed by that PR on main; "(superseded)" one closed by a later operator ruling.

The kiken chain is broken in three places, and all three must land before a withdrawal can be closed on a court: bc-kpnl, bc-wnam, bc-kfup. Update after #444: the Award button works (bc-wnam), and the withdrawn side's next queue row now offers Record fusensho for , which records it and closes the chain on the court. bc-kpnl and bc-kfup still reproduce as filed (re-verified 2026-09-25). Update 2026-10-07: bc-kfup is fixed (#447; a fusenpai on the withdrawn competitor's next bout records Fus.), and bc-kpnl is superseded: the panel was removed by the 2026-09-26 ruling, so each remaining match is closed from its own queue row.

bead	severity	P	title	reviewers	fix direction (top severity)
bc-kfup (closed, #447)	BLOCKS	P1	A fusenpai on a withdrawn competitor's remaining bouts is refused as already ineligible	A	Treat a fusenpai whose loser is already ineligible as a chain award at BOTH guard sites (`checkConcurrentIneligibility` and the K2 check-and-set). Keep the origin CompetitorStatus, keep refusing a second kiken from another match, and list running bouts in the panel.
bc-kpnl (superseded)	BLOCKS	P1	The court console drops the remaining-matches panel as soon as a kiken lands	A, B	Pin the kiken-decided match the way `correctingKey` is pinned until the operator closes the panel, and wire a real close. Do not relax the completed filter.
bc-wnam (closed, #447)	BLOCKS	P1	The remaining-matches panel's Award button throws 'wname is not defined' and records nothing	A	Use the in-scope withdrawn player's name, add a render test that taps Award, and consider no-undef for web-mobile.
bc-htcr (closed, #447)	BLOCKS	P1	A hantei result cannot be corrected: the correction never asks for a reason and the server refuses it	A	Route a correction-mode `submitHantei` through ReasonPrompt with a pending-action ref, and add a render test.
bc-tmfn (closed, #444)	ruling (source B3: SLOWS)	P1	A team encounter can be finished while numbered bouts have no result	B	Finish, client and server, refuses a team match (not kachinuki) while any numbered bout has no result, naming the bouts and never the lineup.
bc-sbq (closed, #447)	SLOWS, high (data loss)	P1	Send back to queue can wipe a reopened encounter's fought bouts, and its confirm misstates what it discards	C, A	Not offered on a match with fought bouts or ReopenPending (End match is the exit). The confirm states what it destroys from current data.
bc-htsd (closed, #462)	SLOWS (BLOCKS while bc-htcr is open)	P2	The hantei SHIRO/AKA wins buttons commit the opposite verdict on a single mis-tap	A	Pick a side, then confirm with a second tap, or separate the buttons.
bc-dsid (closed, #462)	SLOWS, high	P2	The withdrawal and no-show prompt preselects Shiro and records on a bare Record	A, B	No preselected side, Record disabled until one is picked, a button naming the side, and 44px rows.
bc-aadv (closed, #463)	SLOWS	P2	Auto-advance after Finish or a fusenpai loses the next bout's start refusal on the court console	A
bc-cdbl (closed, #463)	SLOWS	P2	Double tap on Call to court sends two announcements	smoke
bc-crpn (closed, #462)	SLOWS	P2	The running match disappears from the court while a correction holds the panel (both orders)	C, A, B
bc-rfsw (closed, #453)	SLOWS, high	P2	A kiken correction silently swaps a team inside the running final	B (J4/26-27)	When a forced correction changes a side of a running downstream match, return that match to the queue or refuse, and tell the court
bc-kcsh (closed, #444)	SLOWS	P2	Correcting a team match decided by kiken shows a blank sheet with no decision or winner	B (J4/23)
bc-dhas (closed, #451)	SLOWS	P2	Adding a representative bout right after the last Tie is erased by that Tie's pending autosave	B (after #444)	Cancel the pending autosave before `onDaihyosen`'s own save, as Finish does.
bc-dhrp (closed, #463)	SLOWS	P2	The daihyosen row offers no way to pick each team's representative	B
bc-dlck (closed, #462)	SLOWS	P2	The decision-locked confirm names neither the started match nor what confirming does	A, B
bc-dtfn (closed, #450)	SLOWS	P2	A double tap passes the two-tap Finish and End match guard	A, B, C
bc-dtip (closed, #450)	SLOWS	P2	A double tap on an ippon button records two ippons	A
bc-emsl (closed, #450)	SLOWS	P2	Tapping an empty mark slot on the team sheet clears the bout's Tie or Fusensho	C
bc-fsnp (closed, #450)	SLOWS	P2	A per-bout fusensho erases the other side's points, and its undo cannot restore them after a reload	B
bc-kosc (closed, #462)	SLOWS	P2	Pools + knockout: knockout bouts are scheduled at the start of the day instead of after the pools	A
bc-lpfb (closed, #458)	SLOWS	P2	The at-court lineup panel shows an empty lineup where the score sheet shows the round default	B
bc-offl (closed, #453)	SLOWS	P2	The court console shows no queued or offline signal after a result is finished offline	A
bc-otpl (closed, #462)	SLOWS	P2	Tapping the Overtime pill opens the glossary instead of the encho counter	A
bc-plcl (closed, #447)	SLOWS	P2	The Pools-tab score editor closes on Start match and after every autosaved point	A
bc-rawm (closed, #444)	SLOWS	P2	Operators see raw server codes, engine messages and internal match ids	A, B, C
bc-rboq (closed, #453)	SLOWS	P2	Record bout offline gives no 'Not sent yet' notice	C
bc-sync (closed, #450)	SLOWS	P2	The sync pill reads Synced while a scoring change is still unsent, and a reload in that window loses it	A, B, C
bc-tmfd (closed, #458)	SLOWS	P2	The team sheet keeps its running total and Record/End controls below the fold on an iPad	B, C
bc-tmwn (closed, #463)	SLOWS (court), COSMETIC (viewer)	P2	The court's Completed list shows no winner for a team result decided by daihyosen, kiken or fusenpai	B
bc-unfx (closed, #467)	SLOWS	P2	An unfought bout in a team encounter is written and shown as a draw	B
bc-cfbd (closed, #458)	SLOWS	P3	A double tap dismisses the confirm dialog it just opened	B
bc-cfcl (closed, #453)	SLOWS	P3	The armed End match label is dark text on a dark fill	C
bc-dscn (closed, #444)	SLOWS	P3	'Discard unsaved scoring changes?' appears where it discards nothing	C, A
bc-flst (closed, #453)	SLOWS	P3	Fighter list: a double tap scores the ippon underneath, and a focused name box will not reopen its list	C
bc-kbrw (closed, #444)	SLOWS (minor)	P3	A double tap on a fought kachinuki bout opens the bout above it	C
bc-lnrf (closed, #458)	SLOWS	P3	The score sheet's lineup refusal renders off screen	B
bc-tp44 (closed, #458)	SLOWS	P3	Operator tap targets under 44px on the score editors and lineup panel	A, B, C
bc-bzlb (closed, #467)	COSMETIC	P3	The 3rd-place match is unlabelled on the court console and missing from Later until both semi-finals finish	A
bc-cpdc (closed, #444)	COSMETIC	P3	Docs say 'Copy from previous match' is at the top of the lineup panel; it is at the bottom	B
bc-kbhn (closed, #444)	COSMETIC	P3	The scoring shortcut hint shows on the iPad and advertises keys the court console ignores	C, A
bc-kheb (closed, #453)	COSMETIC	P3	Kachinuki: one bout's encho marks the whole encounter (E) in the editor header and on the bracket card	C
bc-lbla (closed, #467)	COSMETIC	P3	Operator form labels are not associated with their inputs (login, first-run, wizard fields and pills)	harness
bc-lnul (closed, #458)	COSMETIC	P3	Unsaved lineup entries are lost on reload without a prompt	B, C
bc-lprf (closed, #467)	COSMETIC	P3	The lineup panel's duplicate-member refusal names the wrong position, in three wordings across surfaces	B
bc-rrdr (closed, #463)	COSMETIC	P3	A drawn bout never appears in the public viewer's Recent results	smoke
bc-kfnl (closed, #467)	feature (ruling)	P3	Kachinuki bout mode: an advisory line with fighters left and who is up next	C
bc-kkdc (closed, #444)	docs (ruling)	P3	Docs: describe correcting a kiken recorded against the wrong side	B
bc-tmdc (closed, #467)	docs (ruling)	P3	Docs: team encounter marks appear in the exported workbook, not on the bracket card	B, C
bc-ktih (closed, #467)	not rated	P2	Kachinuki: an unscored bout reads 'Tied bout' after a tie that left one team's last fighter on	#467 mockups
bc-kbrc (closed, #467)	not rated	P2	Kachinuki: after Record bout at exhaustion the bout stays open and nothing shows it was recorded	#467 mockups

50 beads: 4 BLOCKS, 1 ruling bug (P1), 33 SLOWS, 8 COSMETIC, 3 docs or feature from rulings, and 1 review-gap task (bc-shcr, below). A 51st child, bc-mvsj (filed 2026-10-03 on PR #453), asks for a new journey and is not a finding. Two more, bc-ktih and bc-kbrc, were found on 2026-10-11 while making #467's mockups, are added at the end of the table, and closed with #467.

Coverage gap, disclosed (bc-shcr), now closed by the shared-shiaijo journey above: every journey here put each competition on its own shiaijo of a 16-shiaijo tournament, so a shiaijo shared by two competitions was not reviewed in this pass: the competition picker on /admin/shiaijo/:court only renders when a court hosts more than one competition (admin_shiaijo.jsx:1256), and no cross-competition court flow was walked. bc-rfsw, bc-kcsh and the widened bc-crpn were found afterwards, re-examining J4-team-scoring/23 at the operator's request; every step behind that screen is a UI tap (no API writes, no injection).

Rejected after adversarial verification (not filed):

A13, the correction staying pinned after Save correction. This is by design and documented ("use Back to court to return to the live match").
A8's offline-reload part. It duplicates mp-wm4e, a documented gap.
A11's button-colour clause. A red button for an irreversible override matches app convention; the copy half is bc-dlck.
B16, "+ Add new member" minting an extra position while reserves are blank. The guard works as designed per the team-members rules.
C12, the reopen reason prompt preselecting "Ended by mistake", and the matching preset on the individual correction reason. Both are intended and shown that way in the docs.
B15's 26px centre mark slot. It is a recorded layout trade-off.
Tap-target numbers that were a harness artifact. A fullPage screenshot had dropped the coarse pointer; the fix is in shots.mjs. Re-measured with coarse asserted, Lineups Add is 44x44, Save is 100x44 and Bracket Edit result is 44px tall.

Harness notes

Concurrent runs: Playwright empties its outputDir at the start of every run, so two runs in one checkout deleted each other's traces mid-test (ENOENT on context close). E2E_RESULTS_DIR gives a concurrent run its own results folder, and the journey screenshots and audit.json follow it (shots.journeyDir), so two runs never empty each other's folders. It is read from the environment, so every worker resolves the same folder (README).
One server per spec file: Playwright hands a worker one file after another, so a worker-scoped server greeted a second file with the first file's tournament, and the suite passed or failed on the machine's core count. serverForFile starts a fresh server whenever the file changes (found in code review; proven with --workers=1).
fullPage screenshots dropped the coarse pointer: measured, (pointer: coarse) read false after a fullPage shot, which switched off every tap-floor rule for the rest of the test. fixtures/pointer.mjs (restoreCoarse) re-enables touch emulation in place, with no reload, through a CDP session kept for the page's life. Detaching the session clears the emulation it set. Every tap-target measurement (tapSize) asserts a coarse pointer first.
One edit to the capture harness's shared lib (plus a comment wording fix in lib/ui.mjs): scripts/screenshots/lib/editor.mjs gains INLINE_EDITOR, the inline editor's selector, which the e2e fixtures pass as the editor's root. (An optional root argument to its finishMatch was dropped when main's tap-guard version of that function merged in: the journeys finish through their own tapping finishMatch in fixtures/scoring.mjs, and fixtures/pace.mjs imports the library's TAP_DWELL_MS.) Because it touches the capture harness's shared lib, make docs/screenshots SINCE=main selects every capture group on this branch.
Also outside the harness: the Makefile (make e2e, make e2e/deps, E2E_ARGS), .gitignore (scripts/e2e/output/, the harness's package files re-included), and the PR template's new "make e2e green locally" checkbox.
Console policy (fixtures/test.mjs): every context and page a test opens is watched. A console error or an uncaught page error fails the test unless it matches CONSOLE_ALLOWLIST, which holds three entries, each with its reason: a failed /api/ fetch while the test has taken the context offline; the court console's own "Failed to fetch court matches" log while offline; and a 409 on /score or /reopen carrying court_busy, ineligible_competitor or downstream_knockout_running, a refusal the app shows the operator (the court is running another bout, the competitor is fighting elsewhere, or the knockout match it feeds is being fought). Warnings are recorded, not failed. A test whose pages logged anything attaches console.json, and each console error is printed as a CONSOLE {json} line naming the allowlist entry it matched, if any.
Code and docs review (2026-10-12), with sub agents: the journeys had checks that could not fail, now fixed and proved able to fail: bc-tmfn asserts the unscored-bout refusal on each tap; bc-htcr and J2 check that Ht sits in the corrected side's cell; J8 and J2 assert the court moves on after a fusenpai; J1's Scores-tab Next needs two distinct editors on court B; the moved-match, S4a and "nothing started" checks wait for the event rather than a sleep; the individual spec's bead guards are no longer serial with its journeys. The README says where clumsy outcomes are asserted and what "green" counts.
Review rounds on bc-cmpw: round 1: security clean; docs 1 critical (branding logo route public in self-run) and 3 majors, fixed; correctness 1 major (the bc-xrpl probe compared held groups), fixed. Round 2: docs narration rewritten as facts; security 1 major (anyone could complete an unfinished league in self-run), fixed; correctness clean, 1 minor (team-sheet stamp), fixed. Round 3: docs 2 majors (openapi roster wording), fixed; code 1 major (file-mode reset bypassed the single password), resolved by the server-machine-only ruling. Round 4: no critical or major; the Host check and a comment fixed. Harness: two review rounds (infrastructure, journeys, README); round 1's critical (J7 recorded the TV text without asserting) and majors (silent-pass helpers, a too-broad allowlist entry, leaked servers, serial cascades, unasserted results, racy reads) fixed; round 2 had no critical or major, its minors fixed. Antigravity (Gemini) reviewed it independently in three rounds; its verified majors (a refusal body or a late console line failing a test, no server.log on some failures, fixed sleeps in wizardDefaults and bc-ltap) were fixed and round 3 found nothing. The password is called the admin password on every screen; the laptop-shortcut docs state which editors take Left, Right and Esc.
Shared-court seeding (fixtures/shared-court.mjs): two competitions on one shiaijo, created through the wizard and started through the UI, as every other journey.

Files changed

File	What
`.github/pull_request_template.md`	Test plan gains "`make e2e` green locally (any `web-mobile/` change)"
`.gitignore`	Ignore `scripts/e2e/output/`; re-include the harness's `package.json` and `package-lock.json`
`Makefile`	`make e2e` (go/build, then `playwright test` with `E2E_ARGS`) and `make e2e/deps` (npm ci + Chromium, stamped); not in go/test or CI
`scripts/e2e/README.md`	How to run one spec or test, headed mode, layout, the fixme convention, `E2E_RESULTS_DIR`
`scripts/e2e/package.json`	Harness package pinning `@playwright/test` to the same release as `scripts/screenshots`
`scripts/e2e/package-lock.json`	Lockfile for the harness
`scripts/e2e/playwright.config.mjs`	Projects "operator" (iPad Air landscape, coarse) and "public" (phone), Chromium only, output folders
`scripts/e2e/fixtures/test.mjs`	The shared `test`: one server per spec file (`serverForFile`), an operator page whose coarse pointer is verified, and the console policy (`CONSOLE_ALLOWLIST`)
`scripts/e2e/fixtures/pointer.mjs`	The coarse pointer: `isCoarse`, `restoreCoarse` (in place), and `tapSize` for the 44px floor
`scripts/e2e/fixtures/audit.mjs`	Audit rows: the `AUDIT {json}` log line, and `audit.json` written beside each journey's screenshots
`scripts/e2e/fixtures/devices.mjs`	The operator iPad and the spectator phone, shared by projects and hand-made contexts
`scripts/e2e/fixtures/setup.mjs`	First-run tournament form, the real login form, Sign out
`scripts/e2e/fixtures/wizard.mjs`	The create-competition wizard; the one file that knows its copy
`scripts/e2e/fixtures/competition.mjs`	Roster paste box, Generate draw, Discard draw, Start
`scripts/e2e/fixtures/shiaijo.mjs`	`/admin/shiaijo/:court`: the queue, the inline editor and `completedRowFor` (the newest Completed row for a pairing)
`scripts/e2e/fixtures/scores.mjs`	The Scores tab and its overlay editor
`scripts/e2e/fixtures/scoring.mjs`	Scoring an individual bout on either editor mount
`scripts/e2e/fixtures/public.mjs`	The public viewer and the TV board
`scripts/e2e/fixtures/pace.mjs`	The operator's pace: waits out the app's 400ms tap guard (`TAP_DWELL_MS`, imported from the capture harness) after a layer opens or a two-tap button arms
`scripts/e2e/fixtures/clumsy.mjs`	Persona helpers: tapNeighbour, doubleTap, retapWhileSaving, interrupt, hastyConfirm, hastyShiaijoConfirm, and `loudestButton`, the one V4 ranking
`scripts/e2e/fixtures/shots.mjs`	Per-step screenshots; `journeyDir`, the one owner of a journey's output folder; restores the coarse pointer after a `fullPage` shot
`scripts/e2e/fixtures/individual.mjs`	Individual journey helpers: rosters, the decision panel (`recordDecision` takes the side: the prompt picks none), and `correctCompleted` (Correct on a finished bout, sending a live bout back to the queue first when the court refuses)
`scripts/e2e/fixtures/team.mjs`	Team match helpers: naming team members, the three lineup entry points, team bout scoring
`scripts/e2e/fixtures/kachinuki.mjs`	Kachinuki sheet helpers: lineup panel, bout controls, fought rows, correction doors
`scripts/e2e/journeys/smoke.spec.mjs`	Seed through the UI, score on a shiaijo, see it on a public phone; exercises each clumsy helper
`scripts/e2e/journeys/knockout-mixed-individual.spec.mjs`	J1, J2, J6, J8 with their variants, plus the findings' own tests and bc-ltap's (both orders, the save held 150 ms, checked after a reload); J1's reset waits for a removal to land; no fixme left
`scripts/e2e/journeys/knockout-mixed-team.spec.mjs`	J3, J4, J7 with their variants, plus the findings' own tests; no fixme left
`scripts/e2e/journeys/knockout-mixed-kachinuki.spec.mjs`	J5 across seven tests with their variants, plus the findings' own tests; no fixme left
`scripts/e2e/fixtures/shared-court.mjs`	Two competitions on one shiaijo, seeded through the UI; the readers the shared-court journeys share
`scripts/e2e/journeys/shared-shiaijo.spec.mjs`	bc-shcr: two competitions on Shiaijo A (S1 to S9, P1 to P8), plus the fixmes for bc-seqc, bc-pubq, bc-rnrm, bc-pace, bc-sypl and bc-mvrq
`scripts/e2e/journeys/moved-match.spec.mjs`	bc-mvsj: a waiting match moved from court A reaches court B's console without a reload, on two devices; plus the fixmes for bc-stpf and bc-adna
`web-mobile/js/admin_scoring_individual.jsx`	bc-ltap: the board-edit stamp (`boardEditedAtRef`, set by every board edit, cleared when the server's copy is taken) and `adoptServerScore`, which keeps the board over a server copy stamped before the last edit; representative-player picks use the plain autosave call
`web-mobile/js/__tests__/render/individual_editor_tap_back_kept.render.test.jsx`	New: a mark tapped back while its removal is saving stays and is sent; a mark taken off while its add is saving stays off; a change made elsewhere after the last tap is still followed (the first two fail without the fix)
`web-mobile/js/__tests__/render/changed_groups_every_editor.render.test.jsx`	One fixture's server copies now carry a stamp after the tap, as the server's do
`web-mobile/js/admin_shiaijo.jsx`	bc-sdtp: one keyless bounce guard (`startTapRef`, `tapStart`) for every Start match on the court console, the Up next card and the queue rows
`web-mobile/js/__tests__/render/start_match_double_tap.render.test.jsx`	New: a bounce on the card or on a queue row starts nothing; a tap after the window and a keyboard activation start (the first two fail without the fix)
`internal/engine/scoring_tx.go`	bc-xrpl: `ExactScoreReplay` and `answerExactScoreReplay`, an exact resend of a completed score write answered as recorded before any refusal
`internal/engine/exact_score_replay_test.go`	New: the exact-replay rule, 12 cases
`internal/mobileapp/handlers_match.go`	bc-xrpl: `applyCorrectionReason` asks no reason for an exact resend
`internal/mobileapp/score_exact_replay_test.go`	New: an exact resend of a finish is answered as recorded (pool, knockout, with and without named groups; fails without the fix); a same-stamp finish that differs still needs a reason
`web-mobile/js/api_client.jsx`	bc-xrpl: a settled queue entry leaves browser storage at once (`leaveQueue` in `_flushQueue`)
`web-mobile/js/__tests__/queue_ordered_writes.test.jsx`	A sent or refused entry leaves browser storage at once (fails without the change)
`specs/openapi.yaml`	PUT /score: the 400 correction-reason case and the `correctionReason` field describe the exact-resend answer
`docs/architecture/network-architecture.md`	The order-of-writes row: an exact resend of a finish is answered as recorded
`CLAUDE.md`	The correction-reason rule and the queue's replay sentence now agree (exact resend answered as recorded); a settled queue entry leaves storage at once
`internal/mobileapp/confirm_password.go`, `middleware.go`, `server.go`	bc-cmpw: `X-Confirm-Password` check (new file); one password for every organiser route; self-run main-gated list gains the branding logo and league tie-break routes
`internal/mobileapp/handlers_reset.go`	bc-cmpw: file-mode reset only from a loopback peer and Host, else 403 `reset_local_only`
`internal/mobileapp/handlers_competition.go`, `handlers_participants.go`, `handlers_match.go`	bc-cmpw: Delete competition needs the confirm password; completion rules; self-run roster open and check-in writes 409; a participant's exact resend answered as recorded
`internal/mobileapp/auth_admin.go`, `handlers_auth_admin.go` (and their `_test.go`)	bc-cmpw: deleted. The separate admin password and its endpoints
`internal/state/models.go`, `internal/engine/competition.go`, `scoring_tx.go`	bc-cmpw: `CheckInApplies`; `PoolMatchesSettled`, `SwissRoundsToDraw`; the exact-replay probe merges on a copy
`web-mobile/js/api_client.jsx`, `ui.jsx`, `admin_competition_settings.jsx`, `competition_shape.jsx`	bc-cmpw: no admin-password flow; `X-Confirm-Password`; `promptDialog`; the Delete dialog; no check-in in self-run
`web-mobile/js/admin_scoring_autosave.jsx`, `admin_scoring_individual.jsx`, `admin_scoring_team.jsx`	bc-cmpw: `landedAutosave` clears an edit stamp once a later save lands
`web-mobile/js/__tests__/`, `internal//_test.go`	bc-cmpw: tests for each rule above; `admin_elevated.test.jsx` deleted
`docs/user-guide/` (organisers, competitors, commands, install), `docs/architecture/`, `specs/openapi.yaml`, `.claude/rules/match-writes.md`, `CLAUDE.md`, `AGENTS.md`	bc-cmpw: one password, Delete's re-entry, self-run rules, completion rules, reset from the server machine
`scripts/screenshots/lib/editor.mjs`	`INLINE_EDITOR`, the inline editor's selector; a comment says "team", not "fixed-order team"
`scripts/screenshots/lib/ui.mjs`	Comment only: "team", not "fixed-order team" (operator ruling)

Screenshots

The bc-cmpw screenshots:

The bc-ltap fix's screenshots are in its section above (the defect is a lost tap, so the before is the trace and the failing test; there is no layout change). The shared-shiaijo evidence is in its own section too. Below are the knockout and mixed review's evidence shots of the top findings, at operator size (1180x820, coarse pointer) unless noted, captured at 548a936 before main merged #444. Every bead these shots show is now closed on main (bc-kpnl superseded). bc-cdbl's fix (#463) has no e2e test and was not re-checked through the harness. Every step's shot is under scripts/e2e/output/ after make e2e.

Test plan

Closes bc-e2eh
Closes mp-yqxn.2
Closes bc-shcr
Closes bc-mvsj
Closes bc-ltap
Closes bc-sdtp
Closes bc-xrpl
Closes bc-cmpw

🤖 Generated with Claude Code

…nd persona helpers Local gate only (make e2e, make e2e/deps); not in go/test or CI. Reuses scripts/screenshots/lib server/net/editor; finishMatch gains a root argument for the inline shiaijo editor. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Each playwright run empties outputDir at start, so two runs in one checkout deleted each other's traces (ENOENT on context close). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…order team, kachinuki) Persona-audited journeys J1-J8; findings are test.fixme until their beads land. shots.mjs re-applies touch emulation after a fullPage shot, which otherwise drops (pointer: coarse) for the rest of the test. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…urement Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

gitrgoliveira · 2026-09-24T08:21:32Z

2026-10-12: this is the review as performed on 2026-09-24 (at 548a936), with the reviewers' judgement; every finding it names is now closed on main except where the PR body says otherwise. The same journeys' raw rows from a current run are in this comment.

Persona audit A: individual (J1, J2, J6, J8)

Every row is one operator action x clumsy variant, performed in the browser through the harness at 1180x820 with a coarse pointer. Finding ids in this table map to beads as follows: A1 bc-kpnl, A2 bc-kfup, A3 bc-otpl, A4 bc-htcr, A4b bc-htsd, A5 bc-plcl, A6 bc-sbq, A7 bc-kosc, A8 bc-offl (offline reload: mp-wm4e), A9 bc-dtfn, A10 bc-dsid, A11 bc-dlck, A12 bc-rawm, A13 rejected (documented behaviour), A14 bc-dscn, A15 bc-tp44, A16 bc-dtip, A17 bc-aadv, A18 bc-kbhn, A19 bc-sync, A20 bc-offl, A21 bc-bzlb; wname (found in verification) bc-wnam.

Written before the operator ruling that retires the term "fixed-order": read it as "team match" (the team sets its lineup before each match). Tap-target sizes measured before the shots.mjs coarse-pointer fix were re-measured; the bead bodies carry the final numbers.

mp-yqxn.2 reviewer A: persona audit, INDIVIDUAL competitions

Device: operator project, 1180x820, (pointer: coarse) verified. Spec:
scripts/e2e/journeys/knockout-mixed-individual.spec.mjs. Screenshots are relative to
scripts/e2e/output/knockout-mixed-individual/ (last green run). Raw rows as the spec
recorded them: scripts/e2e/output/knockout-mixed-individual/audit.json (gitignored, rewritten each run).

Fixtures (all through the wizard): F1 "Men Individual" pools + knockout, 12 entrants,
shiaijo A and B, prefix M. F1b "Women Individual" pools + knockout, 6, shiaijo E, prefix W
(decisions). F1c "Cadets Individual" pools + knockout, 6, shiaijo F, prefix C (Pools tab).
F2 "Veterans Knockout" knockout only, joint 3rd UNTICKED (bronze match), 4, shiaijo C,
prefix V. F2d "Juniors Knockout" knockout only, default joint 3rd, 4, shiaijo D, prefix J.

J1: shiaijo console, pools to knockout, same court (F1, shiaijo A and B)

journey	step	action	variant	what happened	state correct?	recoverable alone?	taps to recover	undo discoverable?	severity	screenshot
J1	start bout	Start match (Up next card)	correct + size	Bout started, inline editor holds the Up next pairing; button 100x44	yes	n/a	0	n/a	none	J1/01-court-A-idle.png
J1	start bout	Start match (Up next card)	V1 tapNeighbour	not performed: the right neighbour is Call to court, whose double-announcement defect is already known; no scoring state is touched by it	n/a	n/a	n/a	n/a	n/a	n/a
J1	start bout	Start match (Up next card)	V2 doubleTap (18-35ms)	One bout started; second tap landed on "Starting…" (disabled); the next bout moved into Up next and was NOT started	yes	n/a	0	n/a	none	J1/02-start-double-tapped.png
J1	start bout	Start match	V3 / V5	covered by the running-bout interruption rows below	n/a	n/a	n/a	n/a	n/a	n/a
J1	start bout	Start match	V4 hastyConfirm	n/a: no confirm on Start match	n/a	n/a	n/a	n/a	n/a	n/a
J1	award ippon	ippon button	size	58x44	yes	n/a	n/a	n/a	none	n/a
J1	award ippon	Aka M	V1 tapNeighbour right	Hit K on the same side; K recorded	no (wrong waza)	yes	1 (tap the K mark; slot 44x44)	yes: "Tap a scored mark to clear it" shows under the slots while a mark exists	SLOWS	J1/03-ippon-neighbour-tapped.png
J1	award ippon	Aka M	V2 doubleTap (21-57ms)	Two M awarded (M M): the bout reads decided 2-0	no (double award)	yes	1 (tap one mark)	yes (hint)	SLOWS	J1/04-ippon-double-tapped.png
J1	clear mark	tap a scored mark	V2 doubleTap	First tap cleared the mark, second tap was a no-op on the emptied slot	yes	n/a	0	yes	none	J1/05-mark-double-tapped.png
J1	running bout	Shiro K, then reload inside 300ms	V3 interrupt reload (inside the autosave debounce)	Pill read "Synced" at the tap; after the reload the K was gone (outcome varies with timing: once it survived)	no (point lost while the pill said Synced)	yes, from memory	1	n/a	COSMETIC	J1/06-fast-reload-after-tap.png
J1	running bout, one ippon	interruption	V3 reload	Same court A, same match (eyebrow identical), M kept	yes	n/a	0	n/a	none	J1/07-interrupt-reload.png
J1	running bout, one ippon	interruption	V3 hidden (visibilitychange)	Event stream reconnected; same match; M kept	yes	n/a	0	n/a	none	J1/08-interrupt-hidden.png
J1	running bout, one ippon	interruption	V3 back	Back landed on /admin (dashboard); Forward returned to Shiaijo A, same match, M kept	yes	yes	1 (Forward, or the court link)	n/a	none	J1/09-interrupt-back.png
J1	running bout, one ippon	interruption	V3 offline (0.6s)	Same match, M kept, pill back to Synced	yes	n/a	0	n/a	none	J1/10-interrupt-offline.png
J1	running bout	identity after interruptions	V5 lost place	Court name (h1), eyebrow "MEN INDIVIDUAL · POOL A · MATCH 1 OF 3", both names and numbers visible; eyebrow is small caps but the names are large	yes	n/a	0	n/a	none	J1/07-interrupt-reload.png
J1	finish	Finish + Start Next	size	159x44	yes	n/a	n/a	n/a	none	n/a
J1	finish	Finish + Start Next	V2 doubleTap (15-29ms)	The two-tap guard did not hold: the double tap armed AND submitted; bout finished and the next bout started	result right only because the score was already the intended one	yes (Correct)	3-4 (Correct, fix, Save correction, Confirm reason)	Correct is on the Completed row	SLOWS	J1/11-finish-double-tapped.png
J1	finish	armed Finish	V1 tapNeighbour left	not performed: no control within 96px (inline editor has no Cancel; Start match only shows before start). Still armed, court A	n/a	n/a	n/a	n/a	n/a	J1/13-finish-neighbour-tapped.png
J1	finish	armed Finish	V2 retapWhileSaving	Second tap during "Saving…" ignored; exactly one completion; next bout started on A	yes	n/a	0	n/a	none	J1/14-finish-retapped-while-saving.png
J1	finish	Finish	V4 hastyConfirm	n/a: no dialog; the two-tap arm is the guard	n/a	n/a	n/a	n/a	n/a	n/a
J1	finish + start next	the next bout starts	V5 did Finish land?	Only the names change in the editor; the finished result is in the Completed list below the fold or the "Just played" strip; no toast	yes	n/a	0	n/a	COSMETIC	J1/12-bout-2-started-by-finish-next.png
J1	send back	Send back to queue	size	129x44, ghost, below the editor at the page bottom	yes	n/a	n/a	n/a	none	n/a
J1	send back	Send back to queue confirm (a synced M on the board)	V4 hastyShiaijoConfirm	Loud button is the destructive "Send back to queue" (primary); score discarded. In one of two final runs the dialog said "No score has been entered, so nothing will be lost." with the M on the board and the pill Synced	no (copy false; point lost)	yes, from memory	2+ (Start match, re-enter)	n/a	SLOWS	J1/15-send-back-dialog.png, J1/16-send-back-hasty.png
J1	send back	restart the sent-back bout	recovery	Bout back at the top of Up next; restarted with its marks gone	yes	yes	1 + re-entry	yes	none	n/a
J1	send back	Send back to queue	V1 / V2	not performed: the dialog is modal, a second tap lands on the dialog; its neighbours are Cancel (safe)	n/a	n/a	n/a	n/a	n/a	n/a
J1	correct	Correct (Completed row)	size	63x44	yes	n/a	n/a	n/a	none	n/a
J1	correct	Save correction, reason prompt answered by its loud button	V4 hasty (inline prompt)	Confirm with preset "Scoring error" saved "– vs MK"; the panel STAYED on the corrected bout (CORRECTION pill, Save correction still offered)	yes	yes	1 ("← Back to court", page bottom right)	only by scrolling	SLOWS	J1/19-correction-reason.png, J1/20-correction-saved.png
J1	correct	while correcting	V5 lost place	Nothing on screen shows the live bout running on the same court (Up next area gone, editor shows the old bout)	partial	yes	1	only by scrolling	SLOWS	J1/20-correction-saved.png
J1	court B pools	Finish + Start Next through B	correct path	Every bout stayed on B, but Finish + Start Next started QUARTERFINALS MATCH 3 between Pool D bouts: knockout bouts are scheduled at the day's start time (09:00), so a resolved QF jumps the remaining pool bouts	valid but out of order	n/a	0	n/a	SLOWS	J1/22-court-B-pools-done.png
J1	pool -> knockout boundary	Finish + Start Next on A's last pool bout	correct path	Stayed on Shiaijo A and started QF M1	yes	n/a	0	n/a	none	J1/23-boundary-after-last-pool-bout.png
J1	knockout bouts 1-2	Finish + Start Next	correct path	QF M1 then QF M2 then SF M5, all on A	yes	n/a	0	n/a	none	J1/24-court-A-knockout-chained.png
J1	console keys	ArrowRight / ArrowLeft / Esc on the console	correct path (docs: laptop keys)	Nothing moved or closed; the docs (Score a match) say Left/Right move to the previous/next match and Esc closes; the hint row still advertises them on the iPad	no (docs divergence)	n/a	n/a	n/a	COSMETIC	J1/25-console-keys.png
J1	Scores tab nav	Prev / Next	size	Next 62x44; Prev absent on the first row	yes	n/a	n/a	n/a	none	J1/26-scores-overlay-B.png
J1	Scores tab nav	Next x3, ArrowLeft/Right	correct path	All seven editors stayed on Shiaijo B; order interleaves QF and pool bouts (both at 09:00) and walks on into completed bouts	court rule holds	n/a	0	n/a	COSMETIC	J1/27-scores-overlay-after-nav.png
J1	Scores tab nav	Next	V1 tapNeighbour	not performed: the left neighbour is Finish + Start Next, whose single stray tap only arms it	n/a	n/a	n/a	n/a	n/a	n/a

J2: special decisions (F1b shiaijo E; F2 shiaijo C)

journey	step	action	variant	what happened	state correct?	recoverable alone?	taps to recover	undo discoverable?	severity	screenshot
J2	decision controls	Kiken - Voluntary / Kiken - Injury / Fusenpai / Decide by hantei…	size	115x44 / 93x44 / 73x44 / 124x44	yes	n/a	n/a	n/a	none	n/a
J2	kiken	Kiken - Voluntary	V1 tapNeighbour right	Hit Kiken - Injury (25px away): the injury prompt opened, nothing committed	yes	yes	1 (Cancel)	yes	none	J2/01-kiken-neighbour-tapped.png
J2	kiken	side radio "Which side withdrew?"	size	radio 13x13, label 165x18	n/a	n/a	n/a	n/a	SLOWS	J2/02-kiken-prompt.png
J2	kiken	Record, side left on its default	V4 hasty (inline prompt)	Prompt defaults to SHIRO; Aka withdrew; Shiro recorded as withdrawn (Kiken beside the wrong competitor)	no	yes, via the undo below	5 + a forced confirm	Correct is visible; nothing flags the mistake	SLOWS	J2/03-kiken-recorded-default-side.png
J2	kiken chain	remaining-matches panel, 3s later	V3 interrupt (looked away)	The panel was gone before 3s (and, in two runs, before the first read): the court refetch drops the completed bout and unmounts the editor; court idle	no	no: the panel cannot be reopened on the console	n/a	no	BLOCKS	J2/04-kiken-chain-after-3s.png
J2	kiken chain	Start match on the withdrawn competitor's next bout	recovery attempt (console)	Refused with "kiken-voluntary at Pool A-0" (internal id, no advice)	refusal right, copy not	no	n/a	no	SLOWS	J2/05-start-withdrawn-next-bout.png
J2	kiken chain	Fusenpai against the withdrawn competitor (Scores tab overlay, the same call "Award default win" makes)	correct path (the chain)	Refused with the raw code "already_ineligible"; nothing recorded	no	no	n/a	no	BLOCKS	J2/06-chain-fusenpai-scores-tab.png
J2	kiken chain	score M for the opponent and Finish (Scores tab)	recovery attempt	Refused: "kiken-voluntary at Pool A-0"	no	no	n/a	no	BLOCKS	J2/07-chain-scored-instead.png
J2	kiken undo	the court moves on (Up next started; involves the real withdrawer)	correct path	Bout 2 of pool A running	yes	n/a	0	n/a	none	n/a
J2	kiken undo	Correct bout 1 -> Kiken - Voluntary on Aka	V4 hastyConfirm (409 decision_locked)	Dialog "Please confirm": "A subsequent match for one of these competitors has already started... Proceed anyway?"; names neither the match nor what happens to it; loud button is the red "Proceed anyway"; forced: bout 1 now "○○ vs Kiken" (Kiken beside Aka)	intended outcome, but see next rows	yes	1	n/a	SLOWS	J2/09-decision-locked-confirm.png, J2/10-kiken-undo-forced.png
J2	kiken undo	remaining-matches panel after the undo	correct path (the chain)	"No remaining scheduled matches" for Aka: her running bout is not listed	incomplete	no	n/a	no	SLOWS	J2/11-undo-chain-award.png
J2	kiken undo	back to the court	V5 lost place	The bout with the now-withdrawn competitor is still running on E with no marker	no	no (next row)	n/a	no	BLOCKS	J2/12-after-undo-court-E.png
J2	kiken undo	finish that bout (M for the opponent)	recovery attempt	Refused "kiken-voluntary at Pool A-0"; only exit is Send back to queue, where it stays unstartable. Pool A cannot complete	no	no	n/a	no	BLOCKS	J2/14-finish-bout-of-withdrawn.png, J2/13-after-undo-pools-tab.png
J2	fusenpai	Record (Aka did not show)	V2 doubleTap (19-26ms)	One decision: "○○ vs Fus.", no duplicate write, no 409	yes	n/a	0	n/a	none	J2/15-fusenpai-recorded.png
J2	fusenpai	auto-advance to the next bout	correct path	Did NOT advance: the court's Up next was the withdrawn competitor's unstartable bout, the start failed silently, court idle with no message	no	yes	1 (Start on another Upcoming row)	yes	SLOWS	J2/15-fusenpai-recorded.png
J2	fusenpai	Fusenpai	V1 tapNeighbour	not performed separately: its neighbours are the "?" glossary chip and the disabled Fusensho; the Kiken V1 row shows a mis-tap only opens a prompt	n/a	n/a	n/a	n/a	n/a	n/a
J2	knockout tie	Mark draw on a 1-1 knockout bout	correct path	Disabled; explanation only in a title tooltip (unreadable on touch)	yes	n/a	n/a	n/a	COSMETIC	n/a
J2	knockout tie	Finish on a 1-1 knockout bout	V1 (Finish instead of Overtime)	Finish enabled and armable; refused with toast "bracket match m-r1-0: cannot mark completed with no winner; resolve the tie first (daihyosen, or encho on the final kachinuki bout)"; the toast sits on top of the Finish button	yes (not completed)	yes	0	n/a	SLOWS	J2/16-finish-on-knockout-tie.png
J2	encho	Overtime pill	size	110x44	yes	n/a	n/a	n/a	none	n/a
J2	encho	tap the Overtime pill (centre)	correct path	Opened the glossary popover ("延長 · Overtime ...") instead of the encho counter; docs say "open the Overtime control and tick Encho started"	no (docs divergence)	yes	2 (dismiss, tap the timer icon)	no	SLOWS	J2/17-overtime-pill-tapped.png
J2	encho	Encho started checkbox	size	checkbox 13x13, label 165x19	n/a	n/a	n/a	n/a	SLOWS	J2/18-encho-counter.png
J2	encho	+ / -	size	+ 30x44, - 29x44	n/a	n/a	n/a	n/a	SLOWS	J2/18-encho-counter.png
J2	encho	+ x6, double tap +, - to the floor	V2 doubleTap (+)	×7 after six taps, ×9 after the double tap (both counted: unbounded as intended); - floors at ×1 and disables	yes	yes	2 (- twice)	yes	none	J2/18-encho-counter.png
J2	hantei	SHIRO wins / AKA wins	size	87x44 / 72x44, 6px apart	n/a	n/a	n/a	n/a	none	J2/19-hantei-armed.png
J2	hantei	SHIRO wins (intended)	V1 tapNeighbour right	Hit AKA wins: committed at once, no confirm; semi-final completed for the wrong competitor and the bracket advanced	no	no (next row)	n/a	Correct visible, but refused	BLOCKS	J2/21-hantei-neighbour-tapped.png
J2	hantei	the recorded hantei result	correct path	First read "M vs K"; updated in place within 6s to "M (E) K Ht" (docs shape)	yes	n/a	0	n/a	none	J2/20-hantei-result-after-reload.png
J2	hantei	Correct -> SHIRO wins	recovery	No reason prompt; toast "correctionReason: correcting a completed match result requires a non-empty correctionReason"; result unchanged	no	no	n/a	no	BLOCKS	J2/22-hantei-correction-open.png, J2/23-hantei-corrected.png
J2	decisions	any decision	V3 interrupt mid-decision	not performed: the editor interruptions are covered in J1; the decision-specific interruption (looking away after kiken) is the "3s later" row above	n/a	n/a	n/a	n/a	n/a	n/a

J6: confirm + advance OFFLINE (F2, shiaijo C, bronze match on)

journey	step	action	variant	what happened	state correct?	recoverable alone?	taps to recover	undo discoverable?	severity	screenshot
J6	finish offline	the last semi-final running	correct path	Semi-final 2 (Wen Zhao vs Vic Lange) running on C	yes	n/a	0	n/a	none	n/a
J6	finish offline	the screen 300ms after Finish, offline	V5 did Finish land?	Editor gone (it holds the "Not saved yet" banner, so no banner anywhere); header still reads "Connected"; local bracket advanced: final Taro Ono vs Vic Lange, bronze Wen Zhao vs Uma Patel in Up next	locally yes, write queued	n/a	0	n/a	SLOWS	J6/01-offline-finish-300ms.png, J6/02-offline-finished.png
J6	finish offline	the bronze bout in Up next	V5 lost place	Labelled only "10:55 · Veterans Knockout": no "3rd place", no match number; before the semis it was absent from Later	yes	n/a	0	n/a	COSMETIC	J6/02-offline-finished.png
J6	finish offline	Retry now while offline	V2 doubleTap	not performed: no Retry button exists (the banner that carries it never renders)	n/a	n/a	n/a	n/a	n/a	J6/04-offline-retry-double-tapped.png
J6	finish offline	Start the same bout again	V5 lost place	Not offered again (Up next moved to the bronze), so no double entry	yes	n/a	0	n/a	none	J6/05-offline-restart-same-bout.png
J6	finish offline	reload while offline	V3 interrupt reload (offline)	net::ERR_INTERNET_DISCONNECTED, blank page: the console is unusable until the network returns	no (console lost)	yes, when online	1 (reload)	n/a	SLOWS	J6/06-offline-reload.png
J6	reconnect	back online, reopen the court	V5 lost place	The queued Finish (persisted across the failed reload) landed: semi-final "– vs M"; Up next = bronze	yes	n/a	0	n/a	none	J6/07-online-again.png
J6	reconnect	a second page reads the Bracket tab	correct path	Server bracket: final carries the winner, 3rd place match carries the loser	yes	n/a	0	n/a	none	J6/08-server-bracket-after-reconnect.png
J6	finish offline	Finish	V1 / V4	n/a: V1 covered in J1 (no neighbour in reach); no dialog on this path	n/a	n/a	n/a	n/a	n/a	n/a

J8: the same scoring from the Pools tab (F1c) and the Bracket tab (F2d)

journey	step	action	variant	what happened	state correct?	recoverable alone?	taps to recover	undo discoverable?	severity	screenshot
J8	pools tab	match row (opens the editor)	size	376x44	yes	n/a	n/a	n/a	none	J8/01-pools-tab-editor-open.png
J8	pools tab	Start match in the Pools-tab editor	correct path	The editor CLOSED as soon as the start saved	started, but the operator is thrown out	yes	1 (reopen the row)	yes	SLOWS	J8/02-pools-tab-after-start.png
J8	pools tab	one ippon	correct path	The editor closed again ~300ms later (the autosave write closes it); M kept on reopen	point kept, editor lost	yes	1 per point	yes	SLOWS	J8/03-pools-tab-after-ippon.png
J8	pools tab	Finish ("Finish", no Start Next)	V5 lost place	Editor closes; back on the Pools tab with the row reading "– vs M"; nothing points at the next bout (the pool's rows are right there)	yes	n/a	1 (next row)	n/a	COSMETIC	J8/04-pools-tab-after-finish.png
J8	pools tab	✕ Close with a freshly tapped ippon	V4 hastyConfirm	"Please confirm / Discard unsaved scoring changes?" with a red "Discard changes"; chosen, yet the M was still recorded on reopen (autosave had saved it)	no (dialog promises a discard it does not do)	yes	1 (tap the mark)	yes	SLOWS	J8/05-pools-tab-reopened.png
J8	bracket tab	Finish on a semi-final card	V5 lost place	Falls back to the read-only result card with "Edit result"; the next semi is one tap away on its card	yes	n/a	1	n/a	none	J8/07-bracket-tab-after-finish.png
J8	bracket tab	Edit result	size	720x44 (re-measured with (pointer: coarse) asserted; the earlier 720x26 was taken after a fullPage shot had dropped the coarse pointer)	yes	n/a	n/a	n/a	none	J8/07-bracket-tab-after-finish.png
J8	bracket tab	Edit result confirm	V4 hastyConfirm	Loud button "Edit result" (primary) only reopens the editor; nothing changes until Save correction	yes	yes	1 (✕ Close)	yes	none	J8/08-bracket-tab-edit-result.png
J8	bracket tab	second semi-final, default joint 3rd places	correct path	Both semis scored from the Bracket tab; no 3rd place match rendered	yes	n/a	0	n/a	none	J8/09-bracket-tab-semis-done.png
J8	fusenpai (court F)	Record fusenpai on the console	correct path (auto-advance)	On a court with nothing stuck the court moved on to its next bout on F by itself (so the J2 non-advance is the A2 stuck bout plus a silent start failure)	yes	n/a	0	n/a	none	J8/10-fusenpai-auto-advance-F.png
J8	pools / bracket tab	Finish	V2 / V3	not performed here: identical editor body to J1 (same Finish guard, same autosave); J1 rows apply	n/a	n/a	n/a	n/a	n/a	n/a

Row count: 83 (J1 35, J2 28, J6 9, J8 11; header rows excluded).

gitrgoliveira · 2026-09-24T08:21:33Z

2026-10-12: this is the review as performed on 2026-09-24 (at 548a936), with the reviewers' judgement; every finding it names is now closed on main except where the PR body says otherwise. The same journeys' raw rows from a current run are in this comment.

Persona audit B: team, sizes 3 and 5 (J3, J4, J7)

Every row is one operator action x clumsy variant, performed in the browser through the harness at 1180x820 with a coarse pointer. Finding ids in this table map to beads as follows: B1 bc-lpfb, B2 bc-lprf, B3 bc-unfx and bc-tmfn (ruling: every bout is fought), B4 bc-fsnp, B5 bc-sync, B6 bc-dhrp, B7 bc-dsid, B8 bc-kpnl, B9 bc-dlck, B10 bc-dtfn, B11 bc-tmwn, B12 bc-tmfd, B13 bc-lnrf, B14 bc-lprf, B15 bc-tp44 (centre mark slot rejected; Lineups Add/Save re-measured at 44px), B16 rejected (intended), B17 bc-cfbd, B18 bc-lnul, B19a bc-cpdc, B19b bc-tmdc (ruling: narrow the docs), B20 bc-kkdc (ruling: Correct is the sanctioned recovery).

Written before the operator ruling that retires the term "fixed-order": read it as "team match" (the team sets its lineup before each match). Tap-target sizes measured before the shots.mjs coarse-pointer fix were re-measured; the bead bodies carry the final numbers.

Reviewer B audit: fixed-order TEAM competitions (mp-yqxn.2)

Performed in the browser through the harness, operator project (iPad Air 1180x820, (pointer: coarse) verified), spec scripts/e2e/journeys/knockout-mixed-team.spec.mjs, helpers scripts/e2e/fixtures/team.mjs. Every row below was performed and observed (the raw observation is the AUDIT {...} line of the same step in the run's log), except rows marked "not performed" or "n/a". Screenshot paths are relative to scripts/e2e/output/. Judgement columns (discoverable, severity) are mine, formed from the screenshots at operator size.

Fixtures (all through the wizard, roster paste, Generate draw, Start; members named through the Lineups page, the match panel and the score sheet):

F3: "Team Mixed", team, Regular (fixed order), team size 3, Pools + Knockout, 6 teams, shiaijo A + B, prefix M.
F4: "Team KO J3" (shiaijo C, prefix K) and "Team KO J4" (shiaijo D, prefix S): team, Regular, size 5, Knockout only, 4 teams. "Team KO J7" (shiaijo E, prefix F): same, 2 teams (a single final).

Severity: BLOCKS a running tournament / SLOWS the operator / COSMETIC. "Tap target" measurements are bounding boxes under the coarse pointer (floor 44x44); team.tapBox asserts (pointer: coarse) before every measurement (re-measured after the shots.mjs fix d12ad9c).

Row count: 86 rows (J3: 39, J4: 35, J7: 12).

Ruling: three lineup entry points, defect or intentional?

Intentional in design, defective in execution.

Evidence (J3 rows 12, 17, 19, 22, 24, 27, 30):

The three are three LAYERS, not three copies. The Lineups page writes the ROUND default for one team (putTeamLineup(..., round), a "Round" number input); the at-court panel ("Enter lineup" on the Up next card, "Lineup" on each Upcoming row) and the score sheet's name boxes both write the per-MATCH override (putMatchLineup), which wins over the round default. Each serves a different moment: before the day (Lineups page, reachable only from the competition's side navigation, not from the court console), before the encounter at the table (panel), and during the bout (sheet, the only one on screen while scoring). The docs describe exactly this layering ("Each team encounter has its own lineup", "Copy from previous match", "An empty position also shows a box directly on its score sheet row"), so having three is not itself the defect.
They CAN disagree, and did, with no operator error: for a pool match the panel showed all Shiro boxes empty under "Inheriting round default" (row 12, J3-lineups/07-f3-panel-round-default.png) while the Lineups page held Aoki/Baba/Chiba and the score sheet showed Aoki on bout 1. Cause: the panel asks for the round default by the match's exact round index (1 for a pool match) while the Lineups page's "Round 1" is index 0; the score sheet (and viewer/TV) ask with ?fallback=best and find it (FINDING-B1). On a first-round knockout match (index 0) the panel showed the round default correctly (row 27). The knockout FINAL (round index 1) is expected to show the same gap as a pool match; not verified (row 32).
The disagreement turns into data loss through a realistic slip: tapping Shiro's "Save lineup" (the panel shows it empty) writes an EMPTY match override, after which the score sheet's Shiro names vanish for that match while the Lineups page still shows them (row 17, J3-lineups/12-f3-panel-shiro-empty-saved.png, 13-f3-sheet-after-shiro-empty-save.png). Three re-entries on the sheet repaired it (rows 21-24).
Which one a hurried operator reaches first: at the court, "Enter lineup" sits beside "Start match" on the Up next card, so the panel is the first lineup surface they meet; once the match starts the sheet's name boxes are the only one. The Lineups page is never on the court console. So the surface the operator reaches first (the panel) is the one that reads a different layer from the surface everything else reads (the sheet, viewer, TV).
Nothing on the panel or the sheet says which layer a name comes from beyond the small "Inheriting round default" / "Override for this match" label on the panel; the sheet shows no provenance at all.

Fix direction: make the panel read the same resolution the sheet uses (?fallback=best, or better, one shared resolver), and do not write an override for a side whose values did not change.

J3: lineup entry from all three entry points (F3 size 3, F4 size 5)

#	step	action	variant	what happened	state correct?	recoverable alone?	taps to recover	undo discoverable w/o help?	severity	screenshot
1	F3 Lineups page	open, size-3 team	correct	Positions read bare 1, 2, 3; no FIK names (positionsForSize)	yes	n/a	0	n/a	none	J3-lineups/01-f3-lineups-fresh.png
2	EP1 name slot 1	"+ Add new member…", type Aoki, Add	V2 doubleTap (29 ms)	The second tap landed on the confirm's backdrop and dismissed it: 0 dialogs open, nothing named	no (the add silently did not happen)	yes	1 (tap Add again)	yes (Add still visible with the name typed)	COSMETIC	J3-lineups/02-f3-lineups-add-doubletap-dialog.png
3	EP1 name slot 1	"Name slot" confirm	V4 hastyConfirm	Loudest button "Name slot" (primary), message 'Name M2.1 as "Aoki"?'; the loud button is the intended one, and naming a blank slot is reversible (Rename)	yes	n/a	0	n/a	none	J3-lineups/03-f3-lineups-named.png
4	EP1	Add button	tap target (re-measured under asserted coarse pointer)	44 x 44 px (the earlier 44 x 35 was measured after a fullPage shot had dropped the coarse pointer)	n/a	n/a	n/a	n/a	none	J3-lineups/02-f3-lineups-add-doubletap-dialog.png
5	EP1 wrong fighter	pick position 3	V1 wrong option (reserve M2.4)	Wrong member placed; list correctly omits members placed elsewhere (offered: Chiba, M2.4, M2.5)	yes (as picked)	yes	2 (open select, pick Chiba)	yes	none	J3-lineups/03-f3-lineups-named.png
6	EP1 duplicate	type Aoki at 2 (Aoki at 1)	refusal copy	Red alert "Aoki is already at Position 1."; understood, but the page labels the position "1", not "Position 1"	yes	yes	1 (Cancel)	yes	COSMETIC (wording differs per surface: row 16 "at 2", row 22 "at Position 2", row 26 "at Senpo")	J3-lineups/04-f3-lineups-duplicate-refused.png
7	EP1 unsaved lineup	reload before Save	V3 interrupt reload	Position picks silently gone (all three back to "none"), no unsaved-change prompt; the names minted by the confirms survive in Team members	partly (picks lost, members kept)	yes	4 (3 picks + Save)	n/a	COSMETIC	J3-lineups/05-f3-lineups-after-reload-unsaved.png
8	EP1 unsaved lineup	reload	V5 lost place	Back on the same team and round; nothing says the picks were never saved	yes	n/a	n/a	n/a	COSMETIC	J3-lineups/05-f3-lineups-after-reload-unsaved.png
9	EP1 save	Save lineup	V2 retapWhileSaving	"Saving…" seen, re-tap made, one "Lineup saved" toast, lineup correct after reload	yes	n/a	0	n/a	none	J3-lineups/06-f3-lineups-saved.png
10	EP1 save	Save lineup button	tap target (re-measured under asserted coarse pointer)	100 x 44 px (the earlier 100 x 33 was measured after a fullPage shot had dropped the coarse pointer)	n/a	n/a	n/a	n/a	none	J3-lineups/06-f3-lineups-saved.png
11	EP1	any	V3 hidden / offline	not performed: the Lineups page holds no live connection-dependent state beyond the Save; reload (row 7) covers the interruption
12	EP2 open panel	Enter lineup (Up next), pool match	correct	"Inheriting round default" but every Shiro box empty; Lineups page holds Aoki/Baba/Chiba and the sheet shows them (FINDING-B1)	no	no (the operator cannot make the panel show the default)	n/a	no	SLOWS	J3-lineups/07-f3-panel-round-default.png
13	EP2 Aka typed	type Dai / Eto / Fuji	correct	Typed names accepted; position 2's list omits Dai (placed at 1)	yes	n/a	0	n/a	none	J3-lineups/08-f3-panel-aka-typed.png
14	EP2 save Aka	Save lineup (Aka)	V1 tapNeighbour left (71 px)	Hit Shiro's "Copy from previous match": red "No previous match found to copy from." stays on the Shiro side	yes (nothing written)	yes	1 (the intended Save)	yes	COSMETIC (a harmless slip leaves a red error)	J3-lineups/09-f3-panel-v1-neighbour-of-aka-save.png
15	EP2 save Aka	Save lineup (Aka)	correct	"Override for this match"; Save is 86 x 44	yes	n/a	0	n/a	none	J3-lineups/10-f3-panel-aka-saved.png
16	EP2 duplicate	type Dai at 2 (Dai at 1), Save	refusal copy	"Dai is already at 2." names the position being typed, not the one Dai holds (1) (FINDING-B2); both boxes still read Dai, the slot label under 2 reads M1.2	no (wrong copy)	yes	3 (open 2, pick Eto, Save)	partly	SLOWS	J3-lineups/11-f3-panel-duplicate-refused.png
17	EP2 wrong Save	Save lineup (Shiro), boxes shown empty	V1 wrong side	Wrote an empty Shiro match override ("Override for this match"); the score sheet then showed Shiro bouts with NO names although the Lineups page still holds the round default	no	yes	3 (re-enter three names on the sheet)	no (nothing says the round default was overridden)	SLOWS (consequence of B1)	J3-lineups/12-f3-panel-shiro-empty-saved.png, 13-f3-sheet-after-shiro-empty-save.png
18	EP2	panel	V2 / V3 / V4	not performed: the panel has no confirm (V4 n/a); a reload closes the panel (the operator reopens it, 1 tap); double tap on Save was covered on the Lineups page (row 9)
19	EP3 open row 1 list	tap Shiro row 1 name box	correct	List offers M2.1 Aoki … M2.5 (no name yet); it appears only after the team members load (up to ~1 s after the sheet renders)	yes	n/a	0	n/a	COSMETIC	J3-lineups/14-f3-sheet-row1-list.png
20	EP3	name box	tap target	157 x 36 px bar (below 44 tall)	n/a	n/a	n/a	n/a	COSMETIC	J3-lineups/14-f3-sheet-row1-list.png
21	EP3 wrong fighter	pick row 1 Shiro	V1 option below	Baba placed on row 1, slot label moved to M2.2 at once (write is immediate, no Save)	yes (as picked)	yes	2 (open, pick Aoki)	yes	none	J3-lineups/15-f3-sheet-wrong-fighter.png
22	EP3 duplicate	type Baba into row 1 (Baba at row 2)	refusal copy	Refused ("Baba is already at Position 2."), box reverts to Aoki; the message renders in the amber strip BELOW the IV/PW band, off screen while the operator is at row 1, so the typed name just vanishes	yes	yes	0	no (the reason is out of view)	SLOWS	J3-lineups/16-f3-sheet-duplicate.png
23	EP3	name box	V2 / V3 / V4	not performed: a pick writes at once (no Save to double tap, no confirm); interruption behaviour of the sheet is audited in J4
24	three entry points	compare Shiro positions	consistency	After the sheet repair: sheet Aoki/Baba/Chiba = Lineups page M2.1 Aoki/M2.2 Baba/M2.3 Chiba, but on different layers (override vs default)	yes (by value)	n/a	n/a	n/a	see ruling	J3-lineups/17-f3-sheet-final.png
25	F4 Lineups page	open, size-5 team	correct	FIK names Senpo, Jiho, Chuken, Fukusho, Taisho (positionsForSize)	yes	n/a	0	n/a	none	J3-lineups/28-f4-lineups-fik-names.png
26	F4 EP1 duplicate	type Ito at Taisho (Ito at Senpo)	refusal copy	"Ito is already at Senpo." clear	yes	yes	1	yes	none	J3-lineups/29-f4-lineups-duplicate-refused.png
27	F4 EP2 open panel	Enter lineup (semifinal, round index 0)	correct	Panel shows the round default Ito…Ueda (the B1 fallback gap does not bite in round 0)	yes	n/a	0	n/a	none	J3-lineups/30-f4-panel-fik-names.png
28	F4 EP2 pick by number	pick Aka Senpo slot K1.1 (no name)	correct	Slot label K1.1 shows under Senpo; saved	yes	n/a	0	n/a	none	J3-lineups/31-f4-panel-aka-senpo-by-number.png
29	F4 EP3 sheet	type Nakamura into Aka Senpo	correct	Aria-label "Senpo AKA player" (FIK name kept in the label; bout number column shows 1 by the bc-dnst ruling); label stays K1.1	yes	n/a	0	n/a	none	J3-lineups/32-f4-sheet-row1.png
30	F4 EP3 → EP1	name typed on the sheet reaches the Lineups page	consistency	Team members list shows Nakamura once (the blank slot was renamed, not a new member minted)	yes	n/a	0	n/a	none	J3-lineups/33-f4-lineups-aka-member-named-from-sheet.png
31	docs	"Copy from previous match at the top of the lineup panel"	docs check	The button is at the BOTTOM of each side, beside Save lineup	docs diverge	n/a	n/a	n/a	COSMETIC	J3-lineups/07-f3-panel-round-default.png
32	F4 final round	panel on the final (round index 1)	V5	not performed: the J3 F4 final was never reached; J4's final was scored by fusenpai without opening the panel. Expected to show B1 (same exact-round fetch)
33	EP2 Rename	Rename control on the panel	tap target (measured under asserted coarse pointer)	47 x 14 px text link under the position name (admin_schedule_lineup.jsx styles it fontSize 11, height auto, minHeight 0)	n/a	n/a	n/a	n/a	COSMETIC	J3-lineups/11-f3-panel-duplicate-refused.png
34	EP1 add a substitute	"+ Add new member…" at position 3 (Fuji holds slot .3; reserves .4 and .5 are blank)	V4 hastyConfirm	Loudest button "Add member" (primary) on "Add "Goto" as a new member of Team Ryu? This adds a new position to the team. Once added, it can be cleared but never removed."; members 5 → 6. The irreversible choice is the loud one, and it minted a sixth position although two blank reserve slots existed	no (a permanent extra slot the operator did not need)	partly (clear name only, never removed)	n/a	n/a	COSMETIC (the copy names the consequence; the unused reserves are not offered as the alternative)	J3-lineups/18-f3-lineups-add-member-confirm.png, 19-f3-lineups-after-add-member.png
35	pool standings	Finish the F3 pool encounter with bout 1 won, bout 2 drawn, bout 3 never scored	consistency	Sheet showed bout 3 as X before Finish; Pools tab: Team Kame W1 L0 T0 IV1 IL0 IT2 PW1 PL0, Team Ryu IT2: the unscored bout counts as an individual draw (FINDING-B3, standings consequence)	no	no (the operator cannot tell an unfought bout from a draw on the sheet)	n/a	no	SLOWS; BLOCKS where IT decides a pool place	J3-lineups/20-f3-pool-encounter-bout3-unfought.png, 21-f3-pools-standings.png
36	pool kiken chain	Kiken – Voluntary, Shiro (Team Kuma) withdraws from the running pool encounter	correct	"Remaining matches for Team Kuma" appears with "Team Kame · Pool A · Shiaijo A · 09:32 · Award default win to opponent"	yes (for about a second)	n/a	n/a	n/a	none	J3-lineups/22-f3-kiken-remaining-matches.png
37	pool kiken chain	the operator reads the panel before acting	V5 lost place	1.5 s later the panel is gone: the court moved on ("Ready when you are"), Up next is the withdrawn team's remaining match (Kuma vs Kame), and nothing offers the award again (FINDING-B8)	no	partly (see row 39)	n/a	no	SLOWS	J3-lineups/23-f3-kiken-1500ms-later.png
38	pool kiken chain	Award default win to opponent	V1 tapNeighbour / V2 doubleTap	not performed: the award button was no longer on screen (row 37); the Pools tab still shows Kuma vs Kame as unplayed ("vs")						J3-lineups/26-f3-pools-after-kiken-chain.png
39	pool kiken chain	Start match on the withdrawn team's remaining match (Up next)	V5 lost place	Refused, but the only copy is the raw "kiken-voluntary at Pool A-1" (inline and as a toast): it names neither team nor what to do (record a fusenpai / award the default win)	yes (refused)	yes, if the operator knows to open the match and record fusenpai	4+	no	SLOWS	J3-lineups/27-f3-start-withdrawn-team-match.png

J4: fixed-order team scoring at the court (F4, /admin/shiaijo/D)

#	step	action	variant	what happened	state correct?	recoverable alone?	taps to recover	undo discoverable w/o help?	severity	screenshot
1	start SF1	Start match (Up next)	V2 doubleTap (25 ms)	One start, editor holds SF1, Up next moved to SF2, no error; button 100 x 44	yes	n/a	0	n/a	none	J4-team-scoring/01-sf1-started.png
2	start	first strike right after Start	strike within ~200 ms of Start	Observed in a throwaway harness spec (since removed): the sheet opens in PRE-MATCH before the start lands; a strike tapped then shows in its slot, the pill reads Synced, but no write carries it and it is gone after an immediate reload (any later edit would carry it). With writes held 2.5 s the sheet only opens after the start lands, so the window is the fast-network ~200 ms	no	yes (re-enter)	1	no (pill said Synced)	COSMETIC (B5)	none kept
3	bout 1	Shiro kote	V2 doubleTap (22 ms)	Two marks (M K instead of M): the double tap awarded two points	no	yes	1 (tap the K mark)	yes ("Tap a scored mark to clear it" shows above bout 1)	COSMETIC	J4-team-scoring/02-bout1-double-tap-kote.png
4	bout 1	ippon button / mark slot	tap target	Ippon 44 x 44; the centre mark slot the operator taps to clear is 26 x 44	n/a	n/a	n/a	n/a	COSMETIC	J4-team-scoring/02-bout1-double-tap-kote.png
5	bout 2	Aka men	V1 tapNeighbour left (71 px)	Hit bout 2's centre "Tie (hikiwake)": bout 2 marked X	no	yes	1 (tap Tie again)	yes (button reads "✓ Tie")	COSMETIC	J4-team-scoring/03-bout2-neighbour-tap.png
6	running total	after bouts 1-2	correct	IV 1 · PW 1 each, RESULT "DAIHYOSEN"; but bouts 3-5, never touched, already read X / "✓ Tie" (FINDING-B3)	no (unfought bouts shown as draws)	no	n/a	no	SLOWS	J4-team-scoring/04-after-bout2-totals.png
7	running total	legibility	R6 glance (re-measured under asserted coarse pointer)	The IV/PW band sits BELOW bout 5 (band top at 1359 px, viewport 820); bout rows are 205 px tall and 2.4 fit on screen, so the running total is off screen while scoring bouts 1-4	n/a	n/a	n/a	n/a	SLOWS	J4-team-scoring/04-after-bout2-totals.png
8	bout 3	Fusensho Aka, then untoggle	correct	○○ on Aka, untoggle restores Shiro K; Fusensho 76 x 44	yes	n/a	1	yes (title "Click to undo fusensho")	none	J4-team-scoring/05-bout3-fusensho-aka.png
9	bout 3	Fusensho Aka, reload, untoggle	V3 interrupt reload	The fusensho had already erased Shiro's K on the wire (docs: the withdrawing side's points are kept). After reload "✓ Fusensho" ○○ persisted; untoggle clears the flag but leaves ○○ as ordinary points and cannot restore K (FINDING-B4)	no	yes	3 (clear two ○, re-award K, if remembered)	no	SLOWS	J4-team-scoring/06-bout3-after-reload-before-untoggle.png, 07-bout3-untoggle-after-reload.png
10	bout 4	look at a bout not yet fought	V5 which bout is live	Bout 4 reads X with "✓ Tie (hikiwake)" before anyone fought it (FINDING-B3)	no	no	n/a	no	SLOWS	J4-team-scoring/08-bout4-before-it-is-fought.png
11	bout 4	Aka do while offline	V3 interrupt offline	Pill "Offline" while down, "Synced" after; D kept	yes	n/a	0	n/a	none	J4-team-scoring/09-bout4-offline.png
12	bout 4	Aka kote then reload within 300 ms	V3 interrupt reload (instant)	Pill read "Synced" at the reload; after reload only D, the K is gone (B5, judgement: the trigger, an unload inside 300 ms of a tap, is rare)	no	yes	1	no (pill said Synced)	COSMETIC	none (state recorded)
13	bout 4	tab hidden and back	V3 interrupt hidden	Event stream reconnected, same match, D kept	yes	n/a	0	n/a	none	none
14	bout 4	Back then Forward	V3 back + V5 lost place	Back lands on the competition's Scores tab; Forward returns to Shiaijo D with the same encounter open and D kept	yes	yes	1 (Forward)	yes	none	J4-team-scoring/10-bout4-after-back.png
15	bout 5	Tie on the unfought bout 5	correct	Bout 5 already showed "✓ Tie"; the tap turned it OFF (reads vs): the correct action produced the opposite of what the operator meant (FINDING-B3)	no	yes	1 (tap again)	no	SLOWS	J4-team-scoring/11-bout5-tie-tapped.png
16	bout 5	scoring view	R5 / R6 glance	While on bout 5 the page has scrolled: match identity (eyebrow, team names) is off screen, IV/PW band a sliver at the bottom edge	n/a	n/a	n/a	n/a	SLOWS	J4-team-scoring/12-bout5-scoring-view.png
17	tied encounter	all five scored, level	correct	IV 1-1 PW 1-1, "DAIHYOSEN", Finish reads "Needs a winner" and is disabled; the tie-breaker block is highlighted	yes	n/a	0	n/a	none	J4-team-scoring/13-tied-daihyosen-offer.png
18	daihyosen	Add representative bout	correct	DH row appears in place (no reload needed although the inline host's onClose is a no-op); centre (DH); button 153 x 44	yes	n/a	0	n/a	none	J4-team-scoring/14-daihyosen-added.png
19	daihyosen	pick each team's representative	correct	No picker: both sides read "-" (FINDING-B6, docs: "The score editor lets you pick each team's representative from its roster")	no (docs diverge)	no	n/a	n/a	SLOWS	J4-team-scoring/15-daihyosen-row.png
20	daihyosen	Add representative bout	V2 doubleTap	not performed: the button disables itself ("Adding…") and the row replaces the block; one attempt only
21	hantei	SHIRO wins	V1 tapNeighbour right (6 px gap)	Hit "AKA wins": Ht chip on Aka	no	yes	1 (tap SHIRO wins)	yes (both buttons visible, chosen one filled)	COSMETIC	J4-team-scoring/16-hantei-wrong-side.png
22	hantei	SHIRO wins	correct	Ht on Shiro's slot, centre stays (DH) (closed middle respected), RESULT "SHIRO WIN"; buttons 87 x 44	yes	n/a	0	n/a	none	J4-team-scoring/17-hantei-shiro.png
23	finish SF1	Finish + Start Next	V2 doubleTap (18 ms)	The double tap passed the two-tap guard: SF1 finished and SF2 opened in the editor	yes (intended here)	yes (Correct, with a reason)	2+	no	SLOWS (the guard does not stop an impatient double tap)	J4-team-scoring/18-sf1-finished-sf2-open.png
24	finish SF1	did Finish land?	V5	Completed row reads "IV 1–1 / PW 1–1" with no winner mark; the sheet now shows SF2 scrolled mid-page; only the bracket preview shows who won	partly	n/a	n/a	n/a	SLOWS	J4-team-scoring/18-sf1-finished-sf2-open.png
25	SF2 kiken	Kiken – Voluntary	V1 tapNeighbour right (25 px)	Hit "Kiken – Injury": its prompt opened	yes (nothing recorded yet)	yes	1 (Cancel)	yes	none	J4-team-scoring/19-sf2-kiken-neighbour.png
26	SF2 kiken	Record withdrawal (intended: Aka)	V4 hasty (loudest button "Record")	The side picker PRESELECTS Shiro; a hasty Record withdrew the wrong team (Shiro, J4 Tanuki) with a voluntary kiken, which the docs call permanent; the side radios' label rows are 18 px tall	no	yes, but only by the undocumented path in rows 31-34	6	no	SLOWS (high; see B7)	J4-team-scoring/20-sf2-kiken-prompt.png, 21-sf2-kiken-recorded.png
27	SF2 kiken recovery	Correct on the semifinal	V4 recovery	Correct opens a blank team sheet with "Save correction"; nothing on it shows that a kiken is recorded, no reinstate control	no	see row 33	n/a	no	SLOWS	J4-team-scoring/23-sf2-correct-after-kiken.png
28	SF2 kiken	default-win chain (Remaining matches panel)	correct	Not shown: in a knockout the withdrawn team has no scheduled matches left; the chain was exercised in the F3 pool instead (J3 rows 36-39)	n/a	n/a	n/a	n/a	n/a	J4-team-scoring/21-sf2-kiken-recorded.png
29	final fusenpai	Fusenpai (Aka did not show), side picked, on the corrected final (J4 Tanuki vs J4 Inoshishi)	correct	Recorded, no eligibility refusal for the reinstated Tanuki; "Fus." beside the no-show on the bracket, "complete on Shiaijo D"; the Completed rows' result cells are EMPTY for the kiken and fusenpai team results	partly	n/a	n/a	n/a	SLOWS	J4-team-scoring/28-final-fusenpai-prompt.png, 29-final-fusenpai-recorded.png
30	J4	Prev/Next match buttons, retap while Finish saves, encho on a team bout	various	not performed: time; Finish was covered by the double tap (row 23); encho is out of scope for fixed-order ties (daihyosen path)
31	SF2 kiken recovery	the operator starts the final before noticing	V5 lost place	The final (J4 Kitsune vs J4 Inoshishi) started normally; nothing on the court flags the kiken as unusual	yes (as recorded)	n/a	n/a	n/a	none	J4-team-scoring/22-final-started-with-wrong-team.png
32	SF2 kiken recovery	in the correction: Withdrawal disclosure → Kiken – Voluntary → Aka → Record	correct (the kiken-undo)	Accepted after the confirm in row 33	yes	yes	5 taps to get here (Correct, disclosure, Kiken – Voluntary, Aka, Record)	no (not documented; recording-decisions.md calls voluntary kiken permanent)	SLOWS	J4-team-scoring/24-sf2-rekiken-aka-prompt.png
33	SF2 kiken recovery	409 decision_locked confirm	V4 hastyConfirm	Loudest button "Proceed anyway" (danger). Message: "A subsequent match for one of these teams has already started. Overwriting the prior decision now may make those downstream results inconsistent. Proceed anyway?" It does not name the final, nor say that its Shiro team will be swapped. Here proceeding was the intended action	yes	yes	1	n/a	SLOWS (comprehension under time pressure: generic, names no match, no consequence)	J4-team-scoring/25-sf2-decision-locked-confirm.png
34	SF2 kiken recovery	the court after the recovery	V5 lost place	The running final now reads J4 Tanuki vs J4 Inoshishi (the Shiro team silently swapped in the open editor); SF2 bracket shows Kiken beside Kitsune	yes	n/a	n/a	n/a	none	J4-team-scoring/26-sf2-after-rekiken.png, 27-court-after-kiken-recovery.png
35	J4	autosave pill during the 300 ms debounce	R8 feedback	The pill reads "Synced" from the tap until the debounced write goes out (row 12), and in PRE-MATCH before the start lands (row 2)	no (the pill overstates)	n/a	n/a	n/a	COSMETIC	none

J7: one result on every surface (court iPad, head-table iPad, spectator phone, TV board)

#	step	action	variant	what happened	state correct?	recoverable alone?	taps to recover	undo discoverable w/o help?	severity	screenshot
1	bout 1 live	Shiro men at the court	correct	TV board: "IV 1 PW 1", "#1 M vs #1"	yes	n/a	0	n/a	none	J7-one-result/01-tv-after-bout1.png
2	bout 1 live	TV board, bouts 2-5	correct	TV and viewer show X on bouts 2-5 that have not been fought (FINDING-B3, public)	no	no	n/a	n/a	SLOWS	J7-one-result/01-tv-after-bout1.png, 02-viewer-after-bout1.png
3	daihyosen live	Add representative bout at the court	correct	TV shows the DH row with (DH) in its centre	yes	n/a	0	n/a	none	J7-one-result/03-tv-with-daihyosen.png
4	finish elsewhere	head table: hantei Shiro, Finish	two devices	Court console (left open) followed: Completed 1, "complete on Shiaijo E"	yes	n/a	0	n/a	none	J7-one-result/04-head-table-hantei-shiro.png, 05-court-after-head-table-finish.png
5	finish elsewhere	court's Completed row	V5 who won?	Row reads "IV 1–1 PW 1–1", no winner mark	partly	n/a	n/a	n/a	SLOWS	J7-one-result/05-court-after-head-table-finish.png
6	adopt verdict	court opens Correct on the final and leaves it	correct	Court correction editor shows Ht on Shiro	yes	n/a	0	n/a	none	J7-one-result/06-court-correction-open-shiro.png
7	adopt verdict	head table corrects hantei to Aka (reason "Hantei was for Aka")	two devices	Reason prompt required and understood; correction saved	yes	n/a	0	n/a	none	J7-one-result/07-head-table-correction-reason.png, 08-head-table-after-correction.png
8	adopt verdict	court's editor left open	two devices	ADOPTED without a tap: Ht moved to Aka, RESULT "AKA WIN" (CLAUDE.md "Editors adopt the server's verdict" holds for the team daihyosen)	yes	n/a	0	n/a	none	J7-one-result/09-court-correction-after-remote-change.png
9	every surface	viewer Recent results	correct	"IV 1–1 PW 1–1" between the two teams, nothing says Aka won; a knockout win reads like a draw	partly	n/a	n/a	n/a	SLOWS (distinct from the known "drawn match missing from Recent results")	J7-one-result/10-viewer-final.png
10	every surface	viewer Bracket tab	correct	✓ beside J7 Asahi (Aka), matching the corrected result; no Ht mark on the card	yes (winner)	n/a	0	n/a	COSMETIC (docs say marks for a team encounter appear on the bracket)	J7-one-result/12-viewer-bracket-final.png
11	every surface	TV board after completion	correct	"All matches completed", no result shown (by design, nothing wrong)	yes	n/a	0	n/a	none	J7-one-result/11-tv-final.png
12	J7	V1 / V2 / V3 / V4 on the two-device path	variants	not performed here: every control on this path was already probed in J4; J7 is about agreement across surfaces

Findings (severity-ranked; ids are placeholders for the orchestrator)

Functional findings carry a self-contained test.fixme('FINDING-B<n>: ...') at the end of knockout-mixed-team.spec.mjs; each was run once with fixme removed and failed at its docs assertion today. B5 was withdrawn from functional to judgement (its trigger is a synthetic sub-300 ms unload), so the numbering gap is deliberate. B3 has TWO fixmes with one root cause: file one bead.

id	title	severity	kind
B3	An unfought bout is stored as a hikiwake: it reads X / "✓ Tie" on the sheet, the TV board and the viewer, a Tie tap on it turns the draw OFF, and a pool encounter finished with a bout unscored counts it in IT	SLOWS (BLOCKS where IT decides a pool place)	functional (2 fixmes)
B8	On the court console the withdrawn team's "Remaining matches" panel vanishes about a second after a team kiken; the default win is never offered again and the next start is refused with "kiken-voluntary at Pool A-1"	SLOWS	functional
B1	The match lineup panel shows a pool match's round-default lineup as empty ("Inheriting round default"), while the sheet, viewer and TV show it; saving the empty side then wipes those names from the sheet	SLOWS	functional
B7	Kiken / fusenpai prompt preselects Shiro and its button reads a bare "Record": a hasty tap withdraws the wrong team (voluntary kiken, "permanent" per the docs); the only way back is undocumented and the correction sheet shows no kiken	SLOWS (high)	judgement
B4	A per-bout fusensho erases the other side's points already scored (docs: they stay valid); after a reload its undo cannot restore them and leaves the maru as ordinary points	SLOWS	functional
B6	The daihyosen row offers no way to pick each team's representative (docs: "lets you pick each team's representative from its roster")	SLOWS	functional
B2	The panel's "already at" refusal names the position being typed, not the one the member holds ("Dai is already at 2." for Dai at 1)	SLOWS	functional
B9	decision_locked confirm is generic ("A subsequent match for one of these teams has already started ... Proceed anyway?"), names no match and not the swap it causes; the loud button is "Proceed anyway"	SLOWS	judgement
B10	A double tap on Finish passes the two-tap guard (finished SF1 and opened SF2)	SLOWS	judgement
B11	Who won a team encounter decided by daihyosen, kiken or fusenpai is not shown on the court's Completed list ("IV 1–1 PW 1–1" or an empty cell) nor on the viewer's Recent results ("IV 1–1 PW 1–1")	SLOWS	judgement
B12	Team sheet layout at 1180x820: 2.4 bouts per screen (rows 205 px, re-measured under the coarse pointer); the IV/PW band is below bout 5 and the match identity scrolls away while scoring bouts 3-5	SLOWS	judgement
B13	The sheet's lineup refusal renders below the IV/PW band, off screen while the operator is at bout 1; the typed name just vanishes	SLOWS	judgement
B14	Three wordings for one refusal: "at Position 1" (Lineups page), "at 2" (panel), "at Position 2" (sheet); the page labels the position "1"	COSMETIC	judgement
B15	Tap targets below 44 px under the coarse pointer (re-measured with coarse asserted before each measurement): sheet name box 157x36, centre mark slot 26x44, kiken side radio rows 166x18, panel Rename link 47x14. Lineups Add (44x44) and Save (100x44) meet the floor; their earlier 44x35 and 100x33 were a harness artefact	COSMETIC	judgement
B5	The autosave pill reads "Synced" through the 300 ms debounce and in PRE-MATCH before the start lands; a strike in either window is lost if the page unloads before the next edit	COSMETIC	judgement
B16	V4 on "+ Add new member…": the loud button mints a permanent extra position ("can be cleared but never removed") although two blank reserve slots exist	COSMETIC	judgement
B17	A double tap on the Lineups page's Add dismisses its own confirm (second tap lands on the backdrop)	COSMETIC	judgement
B18	Lineups page loses unsaved picks on reload without a prompt	COSMETIC	judgement
B19	Docs: "Copy from previous match at the top of the lineup panel" (it is at the bottom of each side); team-encounter marks "appear on the bracket" but a daihyosen hantei shows no Ht on the bracket card	COSMETIC	docs divergence
B20	Docs say a voluntary kiken is permanent, yet a correction reinstated the team (J4 rows 31-34). Operator decision, not a code fix: that path is B7's only recovery	n/a	docs divergence (decision)

gitrgoliveira · 2026-09-24T08:21:35Z

2026-10-12: this is the review as performed on 2026-09-24 (at 548a936), with the reviewers' judgement; every finding it names is now closed on main except where the PR body says otherwise. The same journeys' raw rows from a current run are in this comment.

Persona audit C: kachinuki, size 5 (J5)

Every row is one operator action x clumsy variant, performed in the browser through the harness at 1180x820 with a coarse pointer. Finding ids in this table map to beads as follows: C1 bc-sbq, C2 bc-crpn, C3 bc-emsl, C4 bc-kheb (ruling: drop the encounter (E) on the header and the bracket card), C5 bc-sync, C6 bc-dtfn, C7 bc-cfcl, C8 bc-tmfd and bc-kfnl (ruling: add an advisory fighters-left line), C9 bc-kbrw, C10 bc-rawm, C11 bc-rawm, C12 rejected (deliberate preset), C13 bc-flst, C14 bc-rboq, C15 bc-kbhn, C16 bc-dscn, C17 bc-tp44, C18 bc-lnul.

Written before the operator ruling that retires the term "fixed-order": read it as "team match" (the team sets its lineup before each match). Tap-target sizes measured before the shots.mjs coarse-pointer fix were re-measured; the bead bodies carry the final numbers.

Reviewer C: J5 kachinuki persona audit (mp-yqxn.2)

Fixture F5, seeded through the UI only: team knockout, kachinuki, team size 5, four teams
(Hayabusa, Kawasemi, Tsubame, Washi), shiaijo A and B, number prefix T. Draw: M1 on A
(Kawasemi v Hayabusa), M2 on B (Washi v Tsubame), final M3 on A. All 20 team members named
by typing into the court console's "Enter lineup" panel.

Device: operator project, 1180x820, (pointer: coarse) verified at every measurement (see
harness defect H1: a fullPage screenshot silently switches the page to a fine pointer; the
spec restores it and records coarseLost).

Spec: scripts/e2e/journeys/knockout-mixed-kachinuki.spec.mjs (7 tests pass, 5 fixme).
Screenshots: scripts/e2e/output/knockout-mixed-kachinuki/NN-*.png
numbered as taken on the last green run. Raw audit records: the AUDIT {...} lines in
the run's log and audit.json in the output folder.

Severity: BLOCKS a running tournament / SLOWS the operator / COSMETIC / ok (no issue).

Audit table (84 rows)

#	step	action	variant	what happened	state correct?	recoverable alone?	taps	undo discoverable?	severity	screenshot
1	seed	Enter lineup (Up next card)	correct path	Lineup panel opens over the console; both teams, five FIK positions each	yes	n/a	0	n/a	ok	01-lineup-A-saved.png
2	seed	Enter lineup	V1	not performed separately: the neighbour of Start match IS Enter lineup, see row 9	-	-	-	-	-	-
3	seed	Enter lineup	V2	not performed: opening a panel twice is idempotent by construction (one modal)	-	-	-	-	-	-
4	seed	Enter lineup	V4	n/a: no confirm	-	-	-	-	-	-
5	M2 lineup	type five names, not saved	V3 reload	Panel closes on reload; all five typed names gone (no draft, no prompt)	yes (nothing was saved)	yes, retype	1 + 5 names + Save	n/a	SLOWS (FINDING-C18)	02-v3-lineup-reload-unsaved.png
6	M2 lineup	type names	V1/V2	n/a: keyboard entry, no tap target to miss	-	-	-	-	-	-
7	M2 lineup	Save lineup	V2 doubleTap (18ms)	One save, one toast "Match lineup saved", names intact	yes	n/a	0	n/a	ok	03-v2-lineup-save-double-tap.png
8	M2 lineup	Save lineup	V4 / V5	n/a: no confirm; interruption covered by row 5	-	-	-	-	-	-
9	M1 start	Start match (Up next)	V1 right	Thumb hits "Enter lineup" (10px away): lineup panel opens instead of starting	yes (nothing changed)	yes, "Close"	1	yes (big Close)	ok	05-v1-start-neighbour.png
10	M1 start	Start match	V2 doubleTap (30ms)	Started once; inline editor on M1 bout 1	yes	n/a	0	n/a	ok	06-v2-start-double-tap.png
11	M1 start	Start match	V3 / V4	n/a: instant action, no confirm	-	-	-	-	-	-
12	M1 bout 1	Shiro ippon M	V2 doubleTap (15ms)	Two M awarded (bout decided 2-0 on one intended point)	no	yes, tap the mark	1	yes: "Tap a scored mark to clear it" shows while a mark exists	SLOWS	08-v2-ippon-double-tap.png
13	M1 bout 1	Shiro ippon M	V1 right	K awarded instead of M (same side, 6px gap)	no (wrong waza)	yes, tap the mark	1	yes	SLOWS	09-v1-ippon-neighbour.png
14	M1 bout 1	scored bout before Record	V3 reload	Same court, same match, same bout; M K kept (autosaved); pairing Fujii v Aoki	yes	n/a	0	n/a	ok	10-v3-reload-bout-1.png
15	M1 bout 1	Record bout	V2 re-tap while "Saving…"	Exactly one bout recorded; next pairing Fujii v Baba (winner stayed on)	yes	n/a	0	n/a	ok	11-v2-record-retap.png
16	M1 bout 2	Record bout	V1 right	Thumb hits End match: it ARMS; its armed label (tap again + "AKA WIN") is dark text on a dark fill, barely readable	yes (only armed)	yes, Record disarms it	0 extra	partly: armed label illegible (FINDING-C7)	SLOWS	12-v1-record-neighbour-end-armed.png
17	M1 bout 2	Record bout	V4	n/a: one tap by design (operator ruling in code), no confirm	-	-	-	-	-	-
18	M1 bout 3	× Remove this bout	V2 doubleTap (20ms)	Appended pairing removed once; bout 2 (Aka 2-0) is the live row again; second tap harmless	yes	yes, Record bout re-appends Goto v Baba	1	yes (Record bout)	ok	13-v2-remove-bout-double-tap.png
19	M1 bout 3	× Remove this bout	V1	not performed: no second control within reach in the tests' layout pass	-	-	-	-	-	-
20	M1 bout 3	× Remove this bout	V4	n/a: no confirm (removes only an unscored bout, per docs)	-	-	-	-	-	-
21	M1 bout 3	Tie (hikiwake)	V2 doubleTap (31ms)	Toggled on then off: no tie, centre back to "vs"	no (net: nothing marked)	yes, one more tap	1	yes (button state visible)	SLOWS	14-v2-tie-double-tap.png
22	M1 bout 3	Tie (hikiwake)	V1 up	Thumb lands on the EMPTY mark slot above the Tie button; the tie is CLEARED (centre X to vs)	no	yes, tap Tie again	1	no: an empty slot is not an undo, nothing says the tie went	SLOWS (FINDING-C3)	15-v1-tie-neighbour.png
23	M1 bout 3	Record bout on a tie	correct path	Both retire, next pair Hara v Chiba; fought row reads X in the centre	yes	n/a	0	n/a	ok	16-bout-4-tied-encho-offered.png
24	M1 bout 4	Encho	V2 doubleTap (30ms)	(E) in the bout centre (right); overtime count x2; header "(E) OVERTIME x2"; no stepper in kachinuki bout mode (removed by operator ruling), so the extra period stays	partly (count wrong)	no (count cannot be reduced in bout mode)	-	no	COSMETIC (part of FINDING-C4)	17-v2-encho-double-tap.png
25	M1 bout 4	encho running	V3 hidden	Event stream resumed; same bout, (E) kept	yes	n/a	0	n/a	ok	18-v3-hidden-encho.png
26	M1 bout 4	Aka scores in encho, Record	correct path	Chiba stays on v Ishii; fought row reads "(E)" with Aka's K	yes	n/a	0	n/a	ok	21-bout-7-open-seven-deep.png
27	M1 bouts 5-8	header after the encho bout	V5	Header keeps "(E) OVERTIME x2" on bouts 5, 6, 7 and 8 (all regulation)	no (false statement)	no	-	-	COSMETIC (FINDING-C4)	23-v5-seven-deep-scrolled-top.png
28	M1 bout 5	Shiro ippon	V2 doubleTap	M M recorded where the truth was M K (operator does not notice, Records)	no	yes, row 36	3	yes (tap the fought row)	SLOWS	21-bout-7-open-seven-deep.png
29	M1 bout 6	Shiro ippon M	V3 offline	Pill "Offline" while down; point shown; resynced to "Synced" on reconnect	yes	n/a	0	n/a	ok	19-v3-offline-bout-6.png
30	M1 bout 6	bout with one point	V3 back (then forward)	Back lands on /admin; Forward returns to shiaijo A, same bout, point kept	yes	n/a	0	n/a	ok	20-v3-back-bout-6.png
31	M1 bout 7	glance seven bouts in	V5 lost place	At scroll top: the live bout's names are visible, Record bout / End match are below the fold; nothing states fighters remaining per side or who is next (IV/PW band suppressed in bout mode by design)	yes	n/a	scroll	n/a	SLOWS (FINDING-C8)	23-v5-seven-deep-scrolled-top.png, 22-bout-7-open-seven-deep-full.png
32	M1 bout 7	tap targets (coarse)	measurement	ippon 44x44, Fusensho 76x44, Tie 92x44, foul + 44x44, Record 106x44, End 96x44, Remove 127x44, Add next 159x44, fought row 642x44, name box 159x54, clear x 44x44, Send back 129x44; mark slot 26x44 (below 44 wide)	-	-	-	-	COSMETIC (FINDING-C17)	-
33	M1 correction	tap fought bout 5	V1 down	Bout 6 opens for correction instead (framed)	yes (nothing changed)	yes, collapse triangle	1	yes (frame + triangle)	ok	24-v1-fought-row-neighbour.png
34	M1 correction	tap fought bout 5	V2 doubleTap (32ms)	First tap opens bout 5, the row grows, the second tap lands on the row ABOVE and bout 4 opens instead (bout 5 closes)	yes (nothing changed) but wrong bout open	yes	2	yes	SLOWS (FINDING-C9)	25-v2-fought-row-double-tap.png
35	M1 correction	collapse triangle	measurement	44x36 under coarse (below 44 tall)	-	-	-	-	COSMETIC (FINDING-C17)	26-bout-5-open-for-correction.png
36	M1 correction	fix bout 5 M M to M K	correct path	Opens in place with full controls, saves as you go; no winner-flip warning (winner unchanged); the live bout 7 keeps its pairing	yes	n/a	3 (open, remove M, add K)	yes	ok	27-bout-5-corrected-open.png
37	M1 correction	bout 5 open while bout 7 is live	V5	Two rows carry full scoring controls at once; the corrected one is framed and has the down triangle, the live one is plain	yes	n/a	-	partly	SLOWS	28-bout-5-and-live-bout-both-editable.png
38	M1 correction	bout 5 open	V3 back	Correction closed on return (component state), the fix was already saved (M K)	yes	n/a	0	n/a	ok	29-v3-back-mid-correction.png
39	M1 correction	collapse triangle	V2 doubleTap (20ms)	Collapsed; did not reopen	yes	n/a	0	n/a	ok	30-v2-collapse-double-tap.png
40	M1 correction	flip bout 1's winner, then restore	correct path	Warning "You changed who won bout 1..." appears; later bouts untouched (per docs); restored cleanly	yes	yes	5	yes	ok	31-bout-1-winner-flipped-warning.png
41	M1 bout 7	Tie on the Aka taisho, Record	correct path	Aka out of fighters; app appends bout 8 Jin v Endo (the fighter who tied kept on, per docs)	yes	n/a	0	n/a	ok	32-bout-7-taisho-tie.png
42	M1 bout 8	Fusensho (Shiro)	V1 left	Thumb hits "H": a hansoku point awarded to Shiro	no	yes, tap the mark	1	yes	SLOWS	33-v1-fusensho-neighbour.png
43	M1 bout 8	Fusensho	V2 doubleTap (27ms)	Ends ON (✓ Fusensho, two maru): the second tap did not toggle it back	yes	n/a	0	n/a	ok	34-v2-fusensho-double-tap.png
44	M1 end	End match (two-tap guard)	V2 doubleTap (26ms)	The double tap COMPLETES the encounter: the guard does not stop a double tap	yes here (intended) but guard defeated	yes, Reopen + reason	2 + reason	yes (Correct, Reopen)	SLOWS (FINDING-C6)	36-v2-end-match-double-tap.png
45	M1 end	Completed row	result	IV 4-2 PW 8-3 (correct); final M3 on A now names Kawasemi	yes	n/a	0	n/a	ok	37-m1-completed.png
46	M1 end	Completed row	V5	In one earlier run the row showed the previous write's IV/PW (3-2 / 6-3) for about a second before settling; not reproduced on the last run	transient	yes	0	n/a	COSMETIC	37-m1-completed.png
47	M1 reopen	Correct (Completed row)	V1 left	No other control within reach (63x44 button)	yes	n/a	0	n/a	ok	38-v1-correct-neighbour.png
48	M1 reopen	Correct	correct path	Finished sheet with all 8 bouts, band "FINAL · 8 BOUTS · SHIRO WIN", Reopen match at the very bottom	yes	n/a	0	n/a	ok	39-m1-correct-completed-view.png
49	M1 reopen	stray tap on the finished sheet	V1	Ippon buttons disabled; nothing changes; no Save correction (by design)	yes	n/a	0	n/a	ok	40-v1-stray-tap-completed-sheet.png
50	M1 reopen	Reopen match	V2 doubleTap (24ms)	One reopen, no error; running with 7 fought rows + bout 8 live	yes	n/a	0	n/a	ok	41-v2-reopen-double-tap.png
51	M1 reopen	Reopen match	V4	n/a: one tap by design (reason owed on the way out, per docs)	-	-	-	-	-	-
52	M1 reopen	reopened encounter	V3 reload	Back on the same encounter, running, same bouts	yes	n/a	0	n/a	ok	42-v3-reload-reopened.png
53	M1 reopen	undo fusensho, award M on bout 8	correct path	Fusensho re-tap restores the pre-walkover score; M awarded; pill "Synced"	yes	n/a	3	yes	ok	43-m1-bout-8-rescored.png
54	M1 reopen	bout 8 correction, 2s idle	V3 reload	Correction persisted (M, no fusensho)	yes	n/a	0	n/a	ok	44-v3-reload-after-bout-8-fix.png
55	M1 reopen	End match -> reason prompt open	V3 reload	Prompt gone, encounter still running, correction kept; End again re-asks	yes	yes	1	yes	ok	46-v3-reload-reason-prompt.png
56	M1 reopen	reason prompt	V4 hasty	Loudest = primary "Confirm" with "Ended by mistake" preselected; the real reason (a scoring error) is recorded wrongly without the operator choosing	partly (audit reason wrong)	no (reason not editable later)	-	-	COSMETIC (FINDING-C12)	47-v4-reason-hasty.png
57	M1 reopen	ended again	result	IV 4-2 PW 7-3, winner still Kawasemi, final keeps its side	yes	n/a	0	n/a	ok	48-m1-ended-again.png
58	M1 earlier run	bout 8 fixed, then End + reload within about 1s	V3 reload	Earlier run only: after the reload bout 8 was back to the walkover, and the ensuing End recorded PW 8-3 (the correction silently lost). Same mechanism as row 83	no	yes if noticed	3	no	SLOWS (FINDING-C5)	(earlier run, overwritten; reproduced as row 83)
59	court busy	move M2 to A (Scores tab court picker)	setup	Organiser step to reach the court-busy door; the Up next card itself has no court picker	yes	n/a	-	-	-	49-m2-moved-to-A.png
60	court busy	Reopen M1 while M2 runs on A	correct path	Panel names M2, warns the requeue clears its score; also shows the raw server line "Court A already has a running match (m-r1-1). Finish that match before reopening this one."	yes	n/a	-	-	COSMETIC (FINDING-C11)	50-m1-reopen-court-busy.png
61	court busy	"Clear its score, queue it, and reopen"	V1 right	Thumb hits "Leave it running" (6px): panel dismissed, nothing changed	yes	yes, Reopen again	1	yes	ok	51-v1-requeue-neighbour.png
62	court busy	court-busy panel	V4 hasty	Loudest = red danger "Clear its score, queue it, and reopen": M2's point is wiped, M2 requeued, M1 reopened. The destructive choice is the prominent one, but its consequence is spelled out above it	as chosen	M2's score: no	-	-	SLOWS	52-v4-requeue-hasty.png
63	court busy	End M1 again, then Start match on Up next (M2)	V5	M2 starts but the panel stays on the finished M1 correction; M2 is nowhere on the page (not in the queue, not in the panel)	no (hidden running match)	yes, "← Back to court" under the sheet	1	no (button below the fold, name does not say M2 is there)	SLOWS (FINDING-C2)	53-v5-m2-started-but-not-shown.png, 54-v5-m2-after-back-to-court.png
64	court busy	M2 restarted	result	Bout 1 empty: its point was cleared, as the panel and the docs say	yes	n/a	-	-	ok	55-m2-restarted-on-A.png
65	M2 bout 2	pick the Shiro fighter (one tap)	correct path	Ueno picked; no stray mark	yes	n/a	0	n/a	ok	57-m2-fighter-picked-one-tap.png
66	M2 bout 2	fighter pick	V2 doubleTap (13ms)	Ueno picked by tap 1; the list closes and tap 2 falls through onto the "D" ippon button underneath: a D point for Shiro	no	yes, tap the mark	1	yes	SLOWS (FINDING-C13)	58-m2-v2-fighter-pick-double-tap.png
67	M2 bout 2	reopen the fighter list	V5	Tapping the name box again does not reopen the list (it still has focus): tap elsewhere, then tap it	-	yes	3	no	SLOWS (FINDING-C13)	58-m2-v2-fighter-pick-double-tap.png
68	M2 bout 2	Esc to close the fighter list (laptop)	V4 hastyConfirm	Esc (list open, pill "Synced") raises "Discard unsaved scoring changes?"; loudest = danger "Discard changes"; after it the pairing is intact (Ueno v Kudo) but the list is STILL open: Esc never closed it	yes	yes, tap elsewhere	1	n/a	COSMETIC (FINDING-C16)	59-m2-esc-discard-dialog.png, 60-m2-after-esc-discard.png
69	M2 bout 2	fighter pick	V1 down	Wada picked instead of Ueno (the option below, 0px gap, 31px tall rows)	no	yes	4	yes (visible name)	SLOWS	61-m2-v1-fighter-pick-neighbour.png
70	M2 bout 3	bout with one point	V3 hidden	Resumed, same bout, point kept	yes	n/a	0	n/a	ok	62-m2-v3-hidden.png
71	M2 bout 4	Record bout	V2 doubleTap (31ms)	Exactly one bout appended (Zaitsu v Kudo); Record disabled on the fresh bout	yes	n/a	0	n/a	ok	63-m2-v2-record-double-tap.png
72	M2 bout 5	Tie on Shiro's taisho, Record	correct path	Taisho stays on v Aka's jiho (per docs table)	yes	n/a	0	n/a	ok	64-m2-taisho-stays-on.png
73	M2 end	End match, taisho defeated	correct path	First tap arms (164x44); armed label illegible again; second tap completes; IV 0-5 PW 0-10	yes	n/a	0	n/a	SLOWS (FINDING-C7)	65-m2-end-armed.png, 66-m2-completed.png
74	M2 reopened	Send back to queue (under a reopened encounter)	V4 hasty	Dialog: "Any score entered on this bout will be discarded"; loudest = primary "Send back to queue"; the encounter's 6-bout log is WIPED, the semifinal restarts at bout 1 and the final loses its side	no	only by re-entering every bout from the paper sheet	~20	no	BLOCKS proposed (FINDING-C1; see the docs note in the findings)	67-m2-send-back-confirm.png, 68-m2-after-send-back.png
75	M2	re-enter and End	result	Re-entered from the sheet, ended with a reason; final regains Tsubame	yes	-	~20	-	-	69-m2-completed-again.png
76	final lineup	Copy from previous match	V2 doubleTap (27ms)	Last run: copied once, 5 names. Earlier run: the rows grew "Rename" links and tap 2 opened Rename on the taisho (Save/Cancel shown, nothing saved)	yes	yes, Cancel	0-1	yes	COSMETIC	70-final-copy-lineup-double-tap.png
77	final lineup	Aka side	V5	Kawasemi's semifinal lineup does not carry over ("Inheriting round default", empty): Copy is needed per side	yes	yes	1	yes	COSMETIC	71-final-lineups-copied.png
78	final running	Reopen M1 while the final it feeds runs on A	correct path	Refused: "cannot reopen: a downstream knockout match has already started or recorded a result" (lowercase server text, no match named, no remedy; the court-busy panel is not offered although the blocker is on this court)	yes (refusal matches docs)	yes	about 6 (Back to court, Send final back, confirm, Correct, Reopen)	no	SLOWS (FINDING-C10)	73-m1-reopen-refused-downstream.png
79	final bout 2	Record bout	V3 offline	Pill "Offline"; the tapped bout stays the live row, Record re-enables, no "Not sent yet" banner; after reconnect the bout is recorded and the next pairing appears	yes (eventually)	yes	0	n/a	SLOWS (FINDING-C14)	74-final-v3-record-offline.png, 75-final-after-reconnect.png
80	final	play out, End	correct path	Tsubame's senpo Kudo wins five; End; knockout complete	yes	n/a	0	n/a	ok	77-final-completed.png
81	final	spectator phone	result	Viewer Recent results: final IV 5-0, M2 IV 0-5, M1 IV 4-2 PW 7-3	yes	n/a	0	n/a	ok	78-viewer-knockout-complete.png
82	all	keyboard hint on the iPad	V5	"M K D T H Shiro · ⇧ Aka · ← → prev/next · Esc close" shows under the coarse pointer although styles.css hides it there	-	-	-	-	COSMETIC (FINDING-C15)	21-bout-7-open-seven-deep.png
83	final bout 4	Shiro M, tab reloads 150ms later	V3 reload	Pill read "Synced" at 150ms; after the reload the point is gone, no warning	no	yes if noticed	1	no	SLOWS (FINDING-C5)	76-final-v3-reload-150ms-after-tap.png
84	all	Call to court	V2	not performed on this journey: known item (admin_shiaijo.jsx callToCourt double announcement)	-	-	-	-	known	-

kachinukiVisiblePositions verdict (admin_scoring_team.jsx:506)

Enough context seven bouts in? For "who fought whom and how": yes. For "where do we
stand": no. kachinukiVisiblePositions returns only the live slot, but the sheet renders
every fought bout above it as a read-only row (kachinukiDoneBoutIdxs, :1611;
renderReadOnlyBout, :2139): bout number, both fighters with member numbers (T2.4 Ishii),
the marks, the centre (vs / X / (E)) and the winner's name in its side colour. Evidence:
22-bout-7-open-seven-deep-full.png, 23-v5-seven-deep-scrolled-top.png. From that list the
operator can work out who stayed on and who retired, but nothing states it:

no per-side "fighters remaining" or "next up" indication. Under exhaustion the operator
must count retired fighters by eye to know End match is due (row 41 was only recognisable
by knowing Endo was the taisho). The IV/PW band is suppressed in bout mode by design
(:3250-3254, user-confirmed brief) and IV/PW does not decide a kachinuki encounter, so
that absence is not the finding (FINDING-C8);
at 1180x820 the sheet is taller than the screen from about bout 6: at scroll top the live
bout's names are in view but Record bout / End match are below the fold (row 31);
the header says "(E) OVERTIME x2" through bouts 5-8 and the bracket card marks the whole
encounter "(E)" (FINDING-C4), which is wrong context.

Can the operator return to a mis-scored earlier bout? Yes, through three doors that work.
(1) Mid-encounter: tap the fought row; it opens in place with full controls, saves as you
go, and a winner flip raises the documented warning (rows 36, 40). Discoverability rests on
a small ▶ in the number column; the only hint on the sheet ("Tap a scored mark to clear it")
does not mention it. A double tap opens the wrong bout (row 34, FINDING-C9). (2) Finished
encounter: Correct, Reopen match (one tap), fix, End match with a reason (rows 48-57); the
fix propagated (PW 8-3 to 7-3, winner and bracket side unchanged). (3) Court busy: the panel
names the running match and offers the atomic requeue (rows 60-64). Remove this bout works
and is recoverable (row 18). The dangerous exit is the one not built for a bout log: "Send
back to queue" under a reopened encounter wipes it (row 74, FINDING-C1). After ending a
reopened correction the panel stays pinned, so the next started match vanishes (row 63,
FINDING-C2).

Who stayed on propagated correctly in every performed case: wins (rows 15, 26), hikiwake
(row 23), a taisho tie under both readings (rows 41, 72), removal and re-record (row 18),
and the bout 8 correction after reopen (row 57).

Findings (severity-ranked; file:line re-derived with grep -n on this worktree)

FINDING-C1 [BLOCKS proposed, judgement: no fixme] "Send back to queue" under a REOPENED kachinuki encounter wipes its whole bout log. admin_shiaijo.jsx:1629 (button, rendered for any running selectedMatch; comment :1617 names it the sanctioned exit for a reopened correction), requestRevert :1136 (scored: scoringStarted(m) :1142), dialog copy :1697 ("Any score entered on this bout"), server POST .../revert-to-queue internal/mobileapp/handlers_match.go:978. Repro: finish M2 (6 bouts), Completed row Correct, Reopen match, tap "Send back to queue" under the sheet, confirm. M2 restarts at bout 1 with nothing recorded; the final loses its side. Evidence 67-m2-send-back-confirm.png, 68-m2-after-send-back.png (row 74). Docs note for the orchestrator: team-tournaments.md promises "Reopening keeps every bout that was fought"; scoring-a-match.md says Send back to queue clears a running match's partial score; a reopened encounter is running, so docs and code agree literally and the spec does not assert it. Fix direction: for a match with fought bouts, do not offer Send back to queue (offer End match / Back to court), or make revert-to-queue refuse a match carrying subResults; at minimum the confirm must count what it destroys ("6 bouts and the result").
FINDING-C2 [SLOWS, functional: test.fixme] After a finished match is pinned for correction, Start match on Up next starts the next match but the panel keeps the correction; the running match is on no part of the page. admin_shiaijo.jsx:806 selectedMatch = correctingMatch || pickedMatch || running[0]; pickMatch :1018 never clears correctingKey (correctMatch :1059, stopCorrecting :1067). Repro: Completed row Correct (or Reopen then End again), then Start match on Up next. Recover: "Back to court" (1 tap, below the sheet). Evidence 53-v5-m2-started-but-not-shown.png, 54-v5-m2-after-back-to-court.png.
FINDING-C3 [SLOWS, functional: test.fixme] Tapping an EMPTY mark slot clears the bout's Tie (and a fusensho). admin_scoring_team.jsx:2994 (slot onClick calls rs.setPts even for an empty slot), setPts :2923/:2951 reset draw/fusensho. Repro: mark Tie on a bout, tap an empty "·" slot beside the centre: X goes back to vs. Evidence 15-v1-tie-neighbour.png.
FINDING-C4 [COSMETIC, functional: test.fixme] Encho on one kachinuki bout marks the whole encounter: header "(E) OVERTIME x2" on every later regulation bout, bracket card "(E)" on the encounter. applyKachinukiEncho admin_scoring_team.jsx:1690 bumps the match-level enchoPeriodCount, enchoBlock :1962 writes it on the wire, eyebrow :2647. CLAUDE.md: in a TEAM match the mark lands on the bout's row, never the encounter. A double tap on Encho doubles the count, with no stepper to undo it in bout mode (operator ruling, not re-litigated). Evidence 17-v2-encho-double-tap.png, 22-bout-7-open-seven-deep-full.png.
FINDING-C5 [SLOWS, functional: test.fixme; autosave is shared, reviewer A may report the same] For the 300ms autosave debounce after a tap the sync pill already reads "Synced"; a reload in that window loses the point with no warning. SyncStatusPill admin_scoring_autosave.jsx:43 tracks the write queue only; markDirty :108 (AUTOSAVE_DEBOUNCE_MS :7) is not reflected. Repro: tap an ippon, reload within ~150ms: point gone. Evidence 76-final-v3-reload-150ms-after-tap.png (row 83; an earlier run lost a whole reopened-bout correction the same way, row 58).
FINDING-C6 [SLOWS, judgement] End match's two-tap guard is defeated by a double tap (26ms): the encounter completes. admin_scoring_team.jsx:3826 arms with no minimum interval before the confirming tap. Recoverable by Correct + Reopen + reason. Evidence 36-v2-end-match-double-tap.png.
FINDING-C7 [SLOWS, judgement] The armed End match label (tap again, then the verdict "AKA WIN") is dark text on the dark confirm fill: the verdict is unreadable at the one moment it must be read. End is btn + btn--confirm (admin_scoring_team.jsx:3817), .btn--confirm (styles.css:562) sets background but no color. Evidence 12-v1-record-neighbour-end-armed.png, 65-m2-end-armed.png.
FINDING-C8 [SLOWS, judgement] Seven bouts deep there is no per-side "fighters remaining / next up" indication, and from about bout 6 Record bout / End match sit below the fold at 1180x820. See the kachinukiVisiblePositions verdict. Evidence 22-..., 23-v5-seven-deep-scrolled-top.png.
FINDING-C9 [SLOWS, judgement] A double tap on a fought bout opens the bout ABOVE it (the row expands under the finger). renderReadOnlyBout admin_scoring_team.jsx:2139 (onClick :2158 -> openDoneBoutEdit :2108). Evidence 25-v2-fought-row-double-tap.png.
FINDING-C10 [SLOWS, judgement] Reopen refused while the downstream final runs shows the raw engine error ("cannot reopen: a downstream knockout match has already started or recorded a result", internal/engine/kachinuki.go:749) at admin_scoring_team.jsx:3643, names no match and offers no remedy, although the blocker is running on this same court (the court-busy panel is not offered). Recovery about 6 taps. Evidence 73-m1-reopen-refused-downstream.png.
FINDING-C11 [COSMETIC, judgement] The court-busy panel prints the raw server line "Court A already has a running match (m-r1-1). Finish that match before reopening this one." (admin_scoring_team.jsx:3665), an internal id and advice that contradicts the button beneath it. Evidence 50-m1-reopen-court-busy.png.
FINDING-C12 [COSMETIC, judgement] The reopen reason prompt preselects "Ended by mistake" (REOPEN_PRESETS admin_scoring_shared.jsx:1152); a hasty Confirm records it even when the reopen was a scoring correction. Evidence 47-v4-reason-hasty.png.
FINDING-C13 [SLOWS, judgement] Fighter pick on a later bout: the list drops over the ippon buttons, so a double tap picks the fighter and the second tap awards the ippon underneath (a D here); and a name box that still has focus does not reopen its list on tap (3 taps). LineupNameInput admin_scoring_shared.jsx:908 (open only onFocus :1026, option commits on mousedown :1052). Evidence 58-m2-v2-fighter-pick-double-tap.png.
FINDING-C14 [SLOWS, judgement] Record bout tapped offline: the pill reads "Offline" but the bout stays the live row, Record re-enables and no "Not sent yet" banner appears (pendingWrite is only set when the host returns {queued:true}, admin_scoring_team.jsx:2496/:3697; the cause of the missing banner was not established). It lands on reconnect. Evidence 74-final-v3-record-offline.png, 75-final-after-reconnect.png.
FINDING-C15 [COSMETIC, judgement] The keyboard shortcut line shows on the iPad: ScoringShortcutHint's inline display: "flex" (admin_scoring_shared.jsx:116) overrides .scoring-shortcut-hint { display: none } under (pointer: coarse) (styles.css:5203). Evidence 21-bout-7-open-seven-deep.png.
FINDING-C16 [COSMETIC, judgement, laptop only] Esc with the fighter list open raises "Discard unsaved scoring changes?" (danger default) while the pill reads Synced (handleDismiss admin_scoring_team.jsx:2584, Esc :2603), and the list stays open afterwards. Evidence 59-m2-esc-discard-dialog.png, 60-m2-after-esc-discard.png.
FINDING-C17 [COSMETIC, judgement] Under the coarse pointer everything measured 44px except the mark slots (26px wide: the "tap a scored mark to clear it" target, admin_scoring_team.jsx:2987) and the collapse triangle (44x36, .tsm-caret-btn styles.css:7372 with margin -10px).
FINDING-C18 [SLOWS, judgement; may overlap reviewer B's J3] Lineup panel: names typed but not saved are lost on reload with no draft or prompt; each side has its own Save. MatchLineupSideEditor admin_schedule_lineup.jsx:80 (values state :111). Evidence 02-v3-lineup-reload-unsaved.png.

Harness defects and things not reachable through the UI

H1 (shared fixtures/shots.mjs, read-only for C): a { fullPage: true } screenshot leaves the operator page at a FINE pointer for the rest of the page's life (measured: coarse true before, false after), so every tap-floor rule switches off and later tap-target measurements are wrong. Reviewers A and B should re-check any measurement taken after a fullPage shot. Workaround in fixtures/kachinuki.mjs ensureCoarse (CDP touch emulation + reload).
H2 (clumsy.hastyConfirm): only answers confirmDialog; the shiaijo page's .shiaijo-move-confirm dialogs, the inline reason prompt and the inline court-busy panel need the local helpers at the end of the spec (candidates for clumsy.mjs).
H3: a LineupNameInput with its list open holds the QUERY as its value and the name as placeholder; boutNames in fixtures/kachinuki.mjs now reads the placeholder in that case (an earlier draft misread a fighter as blanked).
Unreachable: the court-busy requeue door cannot be reached with the natural blocker in a 2-court knockout (the final on the same court is downstream, so the downstream refusal fires first); reaching it needed the Scores tab court picker (the Up next card has none). "+ Add next bout manually" was never needed (the server appended every pairing).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…n and bc-unfx PR #444 changed three behaviours the journeys asserted: - J2: a withdrawn competitor's next bout offers "Record default win" instead of Start match (e78ec2a), and a tied individual knockout bout's Finish reads "Needs a winner", disabled. - J5: ending a reopened encounter asks for no reason (9df3980); the reload interrupt moves to the armed End match. - J3: Copy from previous match moved to the top (bc-cpdc), so nothing sits beside Save lineup; bout 3 is now recorded as a Tie before Finish, per the every-bout-is-fought ruling. bc-tmfn (closed by #444) and bc-unfx (fixed by #444) are plain tests now. New fixme bc-dhas: a representative bout added inside the last Tie's autosave window is erased by that write; J7 and bc-dhrp wait past it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…sult Found by PR #442's kachinuki journey run against this build: Correct on a finished kachinuki encounter, Reopen match, then End match with nothing else changed sent the score write with "changed": [], and the server, told nothing changed, kept the encounter running with no error. useChangedGroups judges a group against the server value the editor last agreed with, seeded when it opened and moved only when it builds a write. A reopen is the server's own write, so the agreed result stayed the finished one, and End rebuilt that same verdict. Once the match the editor renders turns reopenPending, the editor now takes the reopened result as agreed, and that group alone, so every other group keeps its protection against putting back a value another device changed. Two render tests, both seen failing first: End after a reopen names the result; a held edit when the match turns reopened still names no other group (it fails if a reopen re-seeds every group). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…nges - lineupPositionLabel moves to lineup_resolver.jsx, beside alreadyPlacedNote. The at-court panel named a position by its row label ("is already at 1.") where the score sheet and the Lineups page said "is already at Position 1."; all three now use it, and the sheet drops its window lookup and type check. - Changing a position takes a refusal down on the panel and the Lineups page: the refusal was for the lineup as it was, and the next Save judges it again. Found by PR #442's journeys (the refusal stayed up with Save off). - CLAUDE.md records both, and the changed-groups rule for a reopen. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…journeys Run against the build of PR #458's head (51e32fb), not #442's own: all four journeys pass there (18 passed, 17 test.fixme skipped). Lineup UI - bc-lpfb is a plain test: the at-court panel shows the lineup in force with its source line, Shiro's Save lineup is off on a side nothing changed on, and the score sheet names the same fighters after Start match. - J3: Shiro's wrong-side Save becomes "Save lineup is disabled". The V1 neighbour tap now lands on the clear x of Shiro's third name box (the nearest control left of Aka's Save), so the step puts that position back first and records it. - The Lineups page's Round box is the "Lineup for" select: openLineups takes lineupFor (default "Starting lineup"); a new J3 step opens a match's own lineup through it. The unsaved-draft notice is recorded and asserted. - panelSave and typeLineup no longer wait for Save lineup to be enabled again: it stays off until something changes. The saved side is read from its source line. Tap guard (400 ms) - fixtures/pace.mjs holds the dwell. login, hastyConfirm, answerDialog, finishMatch (now local), armFinish, finishTeam, endMatch, awardBoutIppon, the shiaijo confirms and the direct two-tap sites wait it out. Behaviour that moved - Kachinuki: the final carries its teams' lineups (no Copy from previous match); send back to queue keeps the score, so M2 restarts with Aka's M and bout 1 needs only the K; a double tap on Encho can land on Undo encho; a double tap on the first ippon can open the fighter list; a bounce no longer records M M, the mistake is made with a deliberate second tap. - Individual: no remaining-matches panel; the barred notice reads "record the fusensho" and its button is barred-match-record-fusensho. J5 test 4 found a product defect and now asserts the fix (a freshly loaded console lists M1 as completed): End match on a reopened encounter with nothing else changed sent changed:[] and the server left the match running. Fixed in #458 (329000e, useChangedGroups agrees with the reopened result). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…e layer - The team sheet publishes its pinned header's and footer dock's heights (published_height.jsx, now also the owner of the topbar's), and its own content controls carry them as scroll margin: a control a keyboard Tab or a scrollIntoView reaches stops clear of the topbar, the pinned header and the dock (inline), or of the pinned header (overlay); the inline head clears the topbar. Margin on the content, never padding on the scroller: a bar's own buttons lie in a padded strip, and focusing one scrolled the page by about 300 px (measured). Found by PR #442's kachinuki journey, whose double tap on "x Remove this bout" landed on the footer once the held clear-mark line had moved that button 28 px lower. - Escape on an open fighter list closes the list only; the Scores-page score editor used to close with it. - Docs: the two kachinuki captures that show the held spaces, the re-recorded kachinuki demo (chapters unchanged), a sentence on Undo encho's kept place, and the completion step of the round-lineup conversion. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

PR #442's journeys found that the at-court lineup panel, made a dialog in this round, swallowed every tap for the bounce window after it opened, so a Close tapped at once did nothing. The panel is an editor like the overlay score editors, so it now guards only its backdrop against the opening tap, under the same operator ruling (2026-10-06), and its controls work at once; its role, Escape and focus stay. On the Lineups page, the rename box of a member with no name yet was labelled "Rename ", naming nothing to a screen reader; it is labelled with the member's slot, as the rest of the page names it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Conflict in scripts/screenshots/lib/editor.mjs: kept main's tap dwell (TAP_DWELL_MS, sameSideTapPacer), its Finish that also arms engi's Save result and waits out the bounce window, and main's finishMatch signature. Kept the branch's INLINE_EDITOR. The branch's `root` parameter on finishMatch is dropped: no caller passes it, the journeys finish through scripts/e2e/fixtures/scoring.mjs, which taps. fixtures/pace.mjs now imports TAP_DWELL_MS from the library instead of declaring a second 500, and scoring.mjs's note no longer says the library's finishMatch skips the bounce window, since main's waits it out. make e2e: 18 passed, 17 skipped (fixmes), EXIT 0. make go/test: EXIT 0. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…nl's Main now carries #447, #450 and #453. Their beads' fixmes were run unskipped on the merged build: - Plain tests now, passing as written: bc-kfup, bc-htcr, bc-plcl (individual), bc-fsnp, bc-dhas (team), bc-emsl, bc-kheb, bc-sync (kachinuki). - bc-offl: the court header reads "Offline: 1 result not sent" while the Finish is queued (captured offline); the test looked for "Not saved yet". - bc-sbq: #447 made Send back to queue keep the score, and the confirm now says "Any score entered is kept". The test asserts that, then starts the bout again and finds its point still on the board. - bc-kpnl's two tests are removed: the panel they wait for was taken out by the 2026-09-26 ruling (recording a withdrawal changes only that match). Follow-on edits: - The two 800 ms waits that kept bc-dhas from erasing the daihyosen row are gone (J7 and bc-dhrp); bc-dhas's own test proves the fix. - bc-fsnp waits for the sync pill to read Synced instead of 1.5 s: since bc-sync the pill reads Syncing from the first tap. - J5 final plays on from the first unfought bout: bc-kheb, which runs before it in this serial file, records the final's bout 1. Five fixmes remain, all on open beads: bc-otpl, bc-kosc, bc-crpn, bc-lprf, bc-dhrp. make e2e: 28 passed, 5 skipped, EXIT 0. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

With the fixme removed the test passes on main: since 51e32fb (PR #458, "Lineup refusals name a position one way everywhere") the at-court lineup panel refuses a typed duplicate with "Dai is already at Senpo.", composed by lineup_resolver.jsx for every surface. The bead itself is still open. The four fixmes left fail at their own defect assertions with the fixme removed: bc-otpl (no encho checkbox), bc-kosc (a knockout time before the last pool time), bc-dhrp (no name box on the daihyosen row), bc-crpn (no Record bout after Start match, inside the serial flow). make e2e journeys/knockout-mixed-team: 9 passed, 1 skipped, EXIT 0. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…-e2eh-operator-review

gitrgoliveira · 2026-10-12T09:33:03Z

Audit rows, current run: Kachinuki (J5)

Raw rows from make e2e on d3894b4 (2026-10-12, main merged through #467), read from each journey's audit.json. One row per operator action x clumsy variant; the Observed column is what the journey recorded, with screenshot paths under scripts/e2e/output/. These rows carry no judgement: the 2026-09-24 audit comments above are the review as performed, with the reviewers' judgement.

Step	Action	Variant	Observed
M2 lineup	type five names, not saved	V3 interrupt reload	`{"kind":"reload","before":"/admin/shiaijo/B","after":"/admin/shiaijo/B","coarseLost":false,"panelStillOpen":false,"namesKept":["Taki","Ueno","Wada","Yano","Zaitsu"],"draftNotice":"Unsaved lineup changes restored Discard","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/02-v3-lineup-reload-unsaved.png"}`
M2 lineup	Save lineup	V2 doubleTap	`{"gapMs":37,"withinWindow":true,"toasts":["Connected","✅\nMatch lineup saved"],"saved":["Taki","Ueno","Wada","Yano","Zaitsu"],"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/03-v2-lineup-save-double-tap.png"}`
M1 start	Start match (Up next card)	V1 tapNeighbour right	`{"intended":{"tag":"button","label":"Start match","testid":null,"className":"btn btn--primary","disabled":false},"hit":{"tag":"button","label":"Enter lineup","testid":null,"className":"btn btn--sm","disabled":false},"gapPx":10,"lineupOpened":true,"startBox":{"x":46,"y":393.5,"width":100,"height":44},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/05-v1-start-neighbour.png"}`
M1 start	Start match (Up next card)	V2 doubleTap	`{"gapMs":78,"withinWindow":true,"eyebrow":"TEAM KACHINUKI · SEMIFINALS · MATCH 1","running":true,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/06-v2-start-double-tap.png"}`
M1 bout 1	Shiro M	V2 doubleTap	`{"gapMs":46,"withinWindow":true,"marksAfter":["M"],"fighterListOpened":false,"mMovedByPx":0,"recoverTaps":0,"hintShown":true,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/08-v2-ippon-double-tap.png"}`
M1 bout 1	Shiro M (second point)	V1 tapNeighbour right	`{"intended":{"tag":"button","label":"M","testid":null,"className":"ipt-btn ipt-btn--sm ","disabled":false},"hit":{"tag":"button","label":"K","testid":null,"className":"ipt-btn ipt-btn--sm ","disabled":false},"gapPx":6,"marksAfter":["M","K"],"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/09-v1-ippon-neighbour.png"}`
M1 bout 1	scored bout, not yet recorded	V3 interrupt reload	`{"kind":"reload","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","coarseLost":false,"marksKept":["M","K"],"doneRows":0,"pairing":{"shiro":"Fujii","aka":"Aoki"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/10-v3-reload-bout-1.png"}`
M1 bout 1	Record bout	V2 retapWhileSaving	`{"savingSeen":true,"retapped":true,"labelAfter":"Record bout","doneRows":1,"current":{"shiro":"Fujii","aka":"Baba"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/11-v2-record-retap.png"}`
M1 bout 2	Record bout	V1 tapNeighbour right	`{"intended":{"tag":"button","label":"Record bout","testid":null,"className":"btn btn--primary","disabled":false},"hit":{"tag":"button","label":"End match","testid":"kachinuki-end-match-button","className":"btn ","disabled":false},"gapPx":6,"endLabel":"Tap again — AKA WIN","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/12-v1-record-neighbour-end-armed.png"}`
M1 bout 3	× Remove this bout	V2 doubleTap	`{"gapMs":40,"withinWindow":true,"removeBox":{"x":888.078125,"y":697.296875,"width":127,"height":44},"afterRemove":{"doneRows":1,"current":{"shiro":"Fujii","aka":"Baba"},"currentShiroMarks":0,"currentAkaMarks":2},"recoveredTo":{"shiro":"Goto","aka":"Baba"},"recoverTaps":1,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/13-v2-remove-bout-double-tap.png"}`
M1 bout 3	Tie (hikiwake)	V2 doubleTap	`{"gapMs":75,"withinWindow":true,"tieLabel":"Tie (hikiwake)","middle":"vs","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/14-v2-tie-double-tap.png"}`
M1 bout 3	Tie (hikiwake)	V1 tapNeighbour up	`{"intended":{"tag":"button","label":"✓ Tie (hikiwake)","testid":"scoring-modal-tie-button","className":"btn btn--sm btn--primary","disabled":false},"hit":{"tag":"button","label":"·","testid":null,"className":"editor-side__pt ","disabled":false},"gapPx":6,"tieStill":"✓ Tie (hikiwake)","middle":"X","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/15-v1-tie-neighbour.png"}`
M1 bout 4	Encho	V2 doubleTap	`{"gapMs":31,"withinWindow":true,"eyebrow":"TEAM KACHINUKI · SEMIFINALS · MATCH 1","inOvertimeAfterDoubleTap":true,"enchoStillOffered":true,"middle":"(E)","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/17-v2-encho-double-tap.png"}`
M1 bout 4	encho running	V3 interrupt hidden	`{"kind":"hidden","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","resumed":true,"method":"document.hidden override + dispatched visibilitychange","coarseLost":false,"pairing":{"shiro":"Hara","aka":"Chiba"},"middle":"(E)","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/18-v3-hidden-encho.png"}`
M1 bout 5	Shiro M	V2 doubleTap	`{"gapMs":56,"withinWindow":true,"marksAfter":["M"]}`
M1 bout 6	Shiro M	V3 interrupt offline	`{"kind":"offline","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","coarseLost":false,"whileOffline":"Offline","resynced":true,"marks":["M"],"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/19-v3-offline-bout-6.png"}`
M1 bout 6	bout with one point	V3 interrupt back	`{"kind":"back","before":"/admin/shiaijo/A","landedOn":"/admin","hadHistory":true,"after":"/admin/shiaijo/A","coarseLost":false,"pairing":{"shiro":"Ishii","aka":"Doi"},"marks":["M"],"doneRows":5,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/20-v3-back-bout-6.png"}`
harness	fullPage screenshot	measurement	`{"coarseBeforeFull":true,"coarseAfterFull":true,"coarseLost":false}`
M1 bout 7	glance at the sheet seven bouts in (scrolled to top)	V5 lost place	`{"liveBoutNamesInView":true,"recordBoutInView":true,"endMatchInView":true,"runningTotalShown":0,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/23-v5-seven-deep-scrolled-top.png"}`
M1 bout 7	measure tap targets (coarse pointer, floor 44px)	measurement	`{"sizes":{"ippon M":"44x44","Fusensho":"76x44","Tie (hikiwake)":"92x44","foul +":"44x44","empty mark slot":"26x44","Record bout":"106x44","End match":"96x44","× Remove this bout":"127x44","+ Add next bout manually":"159x44","fought bout row":"642x44","fighter name box":"159x46","clear fighter ×":"44x44","Send back to queue":"129x44"},"coarse":true,"tieMinHeight":"44px","tieHeight":"44px"}`
M1 correction	tap fought bout 5 to correct it	V1 tapNeighbour down	{"intended":{"tag":"div","label":"Correct bout 5","testid":"kachinuki-done-bout-4","className":"team-sub-match team-sub-match--readonly team-sub-match--editable","disabled":false},"hit":{"tag":"div","label":"Correct bout 6","testid":"kachinuki-done-bout-5","className":"team-sub-match team-sub-match--readonly team-sub-match--editable","disabled":false},"gapPx":4,"openedForCorrection":{"shiro":"Ishii","aka":"Doi"},"recoverTaps":1,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/24-v1-fought-row-neighbour.png"}
M1 correction	tap fought bout 5 to correct it	V2 doubleTap	`{"gapMs":25,"withinWindow":true,"boutOpenAfter":5,"bout5MarksBefore":{"shiro":["M","M"],"aka":[]},"bout5MarksAfter":{"shiro":["M","M"],"aka":[]},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/25-v2-fought-row-double-tap.png"}`
M1 correction	measure the collapse triangle	measurement	`{"caret":"44x44"}`
M1 correction	tap fought bout 5, fix M M to M K	correct path	`{"winnerFlipWarning":false,"liveBout":{"shiro":"Ishii","aka":"Endo"},"editableRows":2,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/28-bout-5-and-live-bout-both-editable.png"}`
M1 correction	bout 5 open for correction	V3 interrupt back	`{"kind":"back","before":"/admin/shiaijo/A","landedOn":"/admin","hadHistory":true,"after":"/admin/shiaijo/A","coarseLost":false,"stillOpen":0,"bout5":{"shiro":"Ishii","aka":"Chiba"},"bout5Marks":{"shiro":["M","K"],"aka":[]},"live":{"shiro":"Ishii","aka":"Endo"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/29-v3-back-mid-correction.png"}`
M1 correction	collapse triangle	V2 doubleTap	`{"gapMs":14,"withinWindow":true,"openAfter":1,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/30-v2-collapse-double-tap.png"}`
M1 correction	flip bout 1 winner by correction, then restore	correct path	`{"warned":true,"warningText":"You changed who won bout 1. The later bouts were fought based on the old result, so check they still show the right competitors and fix any that are wrong.","liveBoutAfterFlip":{"shiro":"Ishii","aka":"Endo"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/31-bout-1-winner-flipped-warning.png"}`
M1 bout 8	Fusensho (Shiro)	V1 tapNeighbour left	`{"intended":{"tag":"button","label":"Fusensho","testid":"scoring-modal-fusensho-button","className":"btn btn--sm ","disabled":false},"hit":{"tag":"button","label":"H","testid":null,"className":"ipt-btn ipt-btn--sm ipt-btn--h","disabled":false},"gapPx":6,"shiroMarks":["H"],"recoverTaps":1,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/33-v1-fusensho-neighbour.png"}`
M1 bout 8	Fusensho (Shiro)	V2 doubleTap	`{"gapMs":26,"withinWindow":true,"labelAfter":"✓ Fusensho","shiroMarks":["○","○"],"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/34-v2-fusensho-double-tap.png"}`
M1 end	End match (two-tap guard)	V2 doubleTap	`{"gapMs":26,"withinWindow":true,"endBox":{"x":820.03125,"y":768,"width":96,"height":44},"completedByDoubleTap":false,"endLabelAfter":"Tap again — SHIRO WIN","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/36-v2-end-match-double-tap.png"}`
M1 end	Completed row after End match	result	`{"expected":"IV 4–2 PW 7–3","shownFor6s":"IV 4–2 PW 7–3","settledWithoutRefresh":true,"afterRefresh":"IV 4–2 PW 7–3","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/37-m1-completed.png"}`
M1 reopen	Correct (Completed row)	V1 tapNeighbour left	`{"none":"clumsy.tapNeighbour: no control within 96px left of locator('.shiaijo-completed .shiaijo-qrow').filter({ hasText: 'Team Kawasemi' }).filter({ hasText: 'Team Hayabusa' }).last().locator('.shiaijo-row__correct')","correctBox":"63x44","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/38-v1-correct-neighbour.png"}`
M1 reopen	Correct (Completed row)	correct path	`{"boutRowsShown":8,"band":"SHIRO (WHITE) IV: 4 · PW: 7 FINAL · 8 BOUTS SHIRO WIN AKA (RED) IV: 2 · PW: 3","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/39-m1-correct-completed-view.png"}`
M1 reopen	tap an ippon on the completed sheet	V1 stray tap	`{"buttonEnabled":true,"akaMarksBefore":0,"akaMarksAfter":1,"saveButton":0,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/40-v1-stray-tap-completed-sheet.png"}`
M1 reopen	Reopen match	V2 doubleTap	`{"gapMs":52,"withinWindow":true,"reopenBox":{"x":761.03125,"y":768,"width":102,"height":44},"error":"","doneRows":7,"current":{"shiro":"Jin","aka":"Endo"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/41-v2-reopen-double-tap.png"}`
M1 reopen	reopened encounter	V3 interrupt reload	`{"kind":"reload","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","coarseLost":false,"editorBack":true,"eyebrow":"TEAM KACHINUKI · SEMIFINALS · MATCH 1","doneRows":7,"current":{"shiro":"Jin","aka":"Endo"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/42-v3-reload-reopened.png"}`
M1 reopen	undo fusensho, award M on bout 8	correct path	`{"wasFusensho":true,"pill":"Synced","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/43-m1-bout-8-rescored.png"}`
M1 reopen	bout 8 corrected on the live row, 2s idle	V3 interrupt reload	`{"kind":"reload","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","coarseLost":false,"bout8MarksAfterReload":["M"],"fusenshoAfterReload":"Fusensho","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/44-v3-reload-after-bout-8-fix.png"}`
M1 reopen	End match armed	V3 interrupt reload	`{"kind":"reload","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","coarseLost":false,"armedLabel":"Tap again — SHIRO WIN","stillRunning":true,"endedByReload":false,"bout8":{"shiro":"Jin","aka":"Endo"},"bout8Marks":["M"],"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/46-v3-reload-end-armed.png"}`
M1 reopen	End match (two-tap guard)	V2 doubleTap	`{"gapMs":19,"withinWindow":true,"completedByDoubleTap":false,"endLabelAfter":"Tap again — SHIRO WIN","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/47-v2-end-match-double-tap-reopen.png"}`
M1 reopen	ended again: Completed row	result	`{"expected":"IV 4–2 PW 6–3","immediately":"IV 4–2 PW 6–3","settledWithin8s":true,"after":"IV 4–2 PW 6–3","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/48-m1-ended-again.png"}`
M1 reopen, court busy	Correct (Completed row) while M2 runs on A	refused (bc-crpn)	`{"notice":"Shiaijo A is running Match 2 · Team Washi vs Team Tsubame. Finish it or send it back to the queue first. Then correct this match.","toast":"⚠️\nShiaijo A is running Match 2 · Team Washi vs Team Tsubame. Finish it or send it back to the queue first. Then correct this match.\n×","correctionOpened":false,"liveBout":{"shiro":"Taki","aka":"Kudo"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/50-m1-correct-refused-m2-running.png"}`
M1 reopen, court busy	Reopen match (Scores tab) while M2 runs on A	correct path	`{"panel":"Shiaijo A is running Match 2 · Team Washi vs Team Tsubame. Finish it or send it back to the queue first. Sending it back to the queue keeps any score already entered for it: it carries on from there when it is started again. Queue it and reopen Leave it running","requeueBox":"132x44","leaveBox":"109x44","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/51-m1-reopen-court-busy.png"}`
M1 reopen, court busy	Queue it and reopen (Scores tab)	V1 tapNeighbour right	`{"intended":{"tag":"button","label":"Queue it and reopen","testid":"kachinuki-reopen-requeue-button","className":"btn btn--sm btn--danger","disabled":false},"hit":{"tag":"button","label":"Leave it running","testid":"kachinuki-reopen-conflict-dismiss","className":"btn btn--sm btn--ghost","disabled":false},"gapPx":6,"panelStill":false,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/52-v1-requeue-neighbour.png"}`
M1 reopen, court busy	court-busy panel (Scores tab)	V4 hasty (loudest button)	`{"label":"Queue it and reopen","prominence":"danger","otherLabels":["Leave it running"],"m1Reopened":{"shiro":"Jin","aka":"Endo"},"doneRows":7,"upNext":{"shiro":"Team Washi","aka":"Team Tsubame"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/53-v4-requeue-hasty.png","console":"scripts/e2e/output/knockout-mixed-kachinuki/54-v4-requeue-hasty-console.png"}`
M1 reopen, court busy	End M1 again, then Start match on Up next (M2)	V5 lost place	`{"upNextStill":false,"panelShows":"TEAM KACHINUKI · SEMIFINALS · MATCH 2","correctionBadge":0,"m2Live":true,"backToCourt":0,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/55-v5-m2-started-after-m1-ended.png"}`
M1 reopen, court busy	recover the started M2	V5 lost place	`{"recoverTaps":0,"live":{"shiro":"Taki","aka":"Kudo"}}`
M1 reopen, court busy	M2 restarted after the requeue	result	`{"m2Bout1AkaMarks":1,"panelSays":"Sending it back to the queue keeps any score already entered for it: it carries on from there when it is started again","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/56-m2-restarted-on-A.png"}`
M2 bout 2	pick the Shiro fighter (one tap)	correct path	`{"pairing":{"shiro":"Ueno","aka":"Kudo"},"shiroMarksAfterPick":[],"optionBox":"198x31","recoverTaps":0,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/58-m2-fighter-picked-one-tap.png"}`
M2 bout 2	pick the Shiro fighter	V2 doubleTap	`{"gapMs":15,"withinWindow":true,"tapsToReopenList":1,"pairing":{"shiro":"Ueno","aka":"Kudo"},"shiroMarks":[],"recoverTaps":0,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/59-m2-v2-fighter-pick-double-tap.png"}`
M2 bout 2	Esc to close the fighter list	V4 hastyConfirm	`{"dialogShown":false,"editorStill":true,"pairing":{"shiro":"Ueno","aka":"Kudo"},"listStillOpen":false,"pill":"Synced","doneRows":1,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/60-m2-esc-discard-dialog.png","after":"scripts/e2e/output/knockout-mixed-kachinuki/61-m2-after-esc-discard.png"}`
M2 bout 2	pick the Shiro fighter	V1 tapNeighbour down	`{"intended":{"tag":"button","label":"T4.2 Ueno","testid":null,"className":"pmf__option ","disabled":false},"hit":{"tag":"button","label":"T4.3 Wada","testid":null,"className":"pmf__option ","disabled":false},"gapPx":0,"pairingAfterMistap":{"shiro":"Wada","aka":"Kudo"},"shiroMarks":[],"recoverTaps":2,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/62-m2-v1-fighter-pick-neighbour.png"}`
M2 bout 3	bout with one point	V3 interrupt hidden	`{"kind":"hidden","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","resumed":true,"method":"document.hidden override + dispatched visibilitychange","coarseLost":false,"marks":["M"],"pairing":{"shiro":"Wada","aka":"Kudo"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/63-m2-v3-hidden.png"}`
M2 bout 4	Record bout	V2 doubleTap	`{"gapMs":27,"withinWindow":true,"doneRows":4,"current":{"shiro":"Zaitsu","aka":"Kudo"},"recordEnabled":false,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/64-m2-v2-record-double-tap.png"}`
M2 end	End match (first tap arms)	correct path	`{"armed":"Tap again — AKA WIN","endBox":"164x44","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/66-m2-end-armed.png"}`
M2 reopened	Send back to queue (reopened encounter)	V4 hasty confirm	`{"label":"Send back to queue","prominence":"primary","otherLabels":["Cancel"],"title":"Send back to queue?","message":"Team Washi vs Team Tsubame will be returned to the upcoming queue. Any score entered is kept: starting it again carries on from there.","boutsBefore":5,"upNext":{"shiro":"Team Washi","aka":"Team Tsubame"},"boutsAfterRestart":5,"current":{"shiro":"Zaitsu","aka":"Maeda"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/69-m2-after-send-back.png"}`
bc-crpn	Start match (Up next) while M1 is open for correction	refused	`{"notice":"Save the correction of Match 1 · Team Kawasemi vs Team Hayabusa, or leave it with Back to court, then start this match.","toast":"⚠️\nSave the correction of Match 1 · Team Kawasemi vs Team Hayabusa, or leave it with Back to court, then start this match.\n×","screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/71-bc-crpn-start-refused-during-correction.png"}`
final lineup	open the lineup panel for the final	carried over	`{"carried":[{"team":"Team Tsubame","source":"Same as Match 2","lineup":["Kudo","Maeda","Noda","Ono","Sato"]},{"team":"Team Kawasemi","source":"Same as Match 1","lineup":["Fujii","Goto","Hara","Ishii","Jin"]}],"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/72-final-lineups-carried.png"}`
final running	Correct M1 (Completed row) while the final runs on A	refused (bc-crpn)	`{"notice":"Shiaijo A is running Match 3 · Team Tsubame vs Team Kawasemi. Finish it or send it back to the queue first. Then correct this match.","toast":"⚠️\nShiaijo A is running Match 3 · Team Tsubame vs Team Kawasemi. Finish it or send it back to the queue first. Then correct this match.\n×","correctionOpened":false,"liveBout":{"shiro":"Kudo","aka":"Fujii"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/74-m1-correct-refused-final-running.png"}`
final running	Reopen M1 (Scores tab) while its downstream final runs	correct path	`{"message":"Match 3 (Final) is being fought now on Shiaijo A. Finish it or send it back to the queue, then reopen this match again.","courtBusyPanel":0,"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/75-m1-reopen-refused-downstream.png"}`
final bout 2	Record bout	V3 interrupt offline	`{"kind":"offline","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","coarseLost":false,"whileOffline":{"doneRows":1,"banner":"Not sent yet: saved on this device, and sent when the connection returns.","recordLabel":"Record bout","live":{"shiro":"Kudo","aka":"Goto"}},"appendedAfterReconnect":true,"doneRows":2,"live":{"shiro":"Kudo","aka":"Hara"},"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/77-final-after-reconnect.png"}`
final bout 4	Shiro M, then the tab reloads 150ms later	V3 interrupt reload	`{"kind":"reload","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","coarseLost":false,"pillAt150ms":"Syncing…","marksAfterReload":["M"],"screenshot":"scripts/e2e/output/knockout-mixed-kachinuki/78-final-v3-reload-150ms-after-tap.png"}`

gitrgoliveira · 2026-10-12T09:33:04Z

Audit rows, current run: Shared shiaijo (bc-shcr: S1 to S9, P1 to P8)

Raw rows from make e2e on d3894b4 (2026-10-12, main merged through #467), read from each journey's audit.json. One row per operator action x clumsy variant; the Observed column is what the journey recorded, with screenshot paths under scripts/e2e/output/. These rows carry no judgement: the 2026-09-24 audit comments above are the review as performed, with the reviewers' judgement.

Journey	Step	Action	Variant	Observed
S1-S8-S7	S8 slots	wizard default Start times (both created before any roster)	P6 record	`{"si":{"startTime":"09:00","sharedCourtHint":"Concurrency for this competition equals the number of shiaijo (courts) assigned. Different competitions can share shiaijo (courts); the schedule prevents conflicts."},"st":{"startTime":"09:30","sharedCourtHint":"Concurrency for this competition equals the number of shiaijo (courts) assigned. Different competitions can share shiaijo (courts); the schedule prevents conflicts."}}`
S1-S8-S7	S8 slots	every slot on Shiaijo A after Start	P6 record (D3)	{"slotsOnA":[{"comp":"Shared Individual","match":"Pool A · Match 1","time":"09:00"},{"comp":"Shared Individual","match":"Pool A · Match 2","time":"09:05"},{"comp":"Shared Individual","match":"Pool A · Match 3","time":"09:10"},{"comp":"Shared Individual","match":"Pool B · Match 1","time":"09:15"},{"comp":"Shared Individual","match":"Pool B · Match 2","time":"09:20"},{"comp":"Shared Individual","match":"Pool B · Match 3","time":"09:25"},{"comp":"Shared Individual","match":"M1","time":"09:30"},{"comp":"Shared Team","match":"M1","time":"09:30"},{"comp":"Shared Individual","match":"M2","time":"09:35"},{"comp":"Shared Individual","match":"M5","time":"09:40"},{"comp":"Shared Individual","match":"M7","time":"09:45"},{"comp":"Shared Team","match":"M2","time":"09:46"},{"comp":"Shared Team","match":"M3","time":"10:02"}],"sameTimeOnA":[{"comp":"Shared Individual","match":"M1","time":"09:30"},{"comp":"Shared Team","match":"M1","time":"09:30"}],"siKnockout":[{"match":"M1","court":"A","time":"09:30"},{"match":"M2","court":"A","time":"09:35"},{"match":"M3","court":"B","time":"09:30"},{"match":"M4","court":"B","time":"09:35"},{"match":"M5","court":"A","time":"09:40"},{"match":"M6","court":"B","time":"09:40"},{"match":"M7","court":"A","time":"09:45"}],"stKnockout":[{"match":"M1","court":"A","time":"09:30"},{"match":"M2","court":"A","time":"09:46"},{"match":"M3","court":"A","time":"10:02"}]}
S1-S8-S7	S1 picker	open /admin/shiaijo/A	correct path	`{"officiating":"Shared Individual","options":["Shared Individual","Shared Team"],"why":"no bout running: the default is the first competition with a scheduled match on the court (effectiveCompId)","upNext":{"head":"09:00 · Shared Individual · Match 1 of 3","names":["Eri Kato","Aoi Sato"]},"nudges":[{"text":"2 Shared Team matches also waiting on this court.","quiet":true}],"shot":"scripts/e2e/output/shared-shiaijo/04-s1-court-A-picker.png"}`
S1-S8-S7	S1 picker	competition picker	V1 tapNeighbour (left, else down)	`{"direction":"down","hit":{"tag":"button","label":"2 Shared Team matches also waiting on this court. Switch to Shared Team","testid":"shiaijo-nudge","className":"alert shiaijo-nudge shiaijo-nudge--also","disabled":false},"gapPx":12,"url":"/admin/shiaijo/A","officiating":"Shared Team","shot":"scripts/e2e/output/shared-shiaijo/05-s1-picker-neighbour-tapped.png"}`
S1-S8-S7	S1 picker	competition picker	size	`{"w":213,"h":53,"meets44":true}`
S1-S8-S7	S1 picker	open /admin/shiaijo/B (SI alone)	correct path	`{"officiating":"Shared Individual","picker":false}`
S1-S8-S7	S8 slots	Auto-schedule competition: Shared Team from 09:12	P6 record (D3)	`{"spacing":"3:00","stAfter":[{"match":"M1","court":"A","time":"09:12"},{"match":"M2","court":"A","time":"09:15"},{"match":"M3","court":"A","time":"10:02"}],"siUnchanged":true}`
S1-S8-S7	S7 public	TV board A and the public schedule before play	P8 record (D2)	{"tv":"Shared Court Cup · SHIAIJO A Shared Individual POOL A · 0 / 3 MATCHES SI2 Eri Kato vs Aoi Sato SI1 SI3 Jun Oda vs Aoi Sato SI1 SI3 Jun Oda vs Eri Kato SI2 UP NEXT Pool B 3 bouts SI5Hana ItovsChie MoriSI4 SI6Lea RothvsChie MoriSI4 SI6Lea RothvsHana ItoSI5","schedule":"← SHARED COURT CUP Schedule All matches across courts and competitions Filter by player, team, dojo or number… All competitions Shared Individual Shared Team All Shiaijo Shiaijo A Shiaijo B 14 matches of 14 12 Oct 2026 SHIAIJO A 8 matches 09:00 POOL A Next up Shiro: SI2 Eri Kato Aka: SI1 Aoi Sato Shared Individual 09:05 POOL A #2 Shiro: SI3 Jun Oda Aka: SI1 Aoi Sato Shared Individual 09:10 POOL A #3 Shiro: SI3 Jun Oda Aka: SI2 Eri Kato Shared Individual 09:12 SEMIFINALS Next up Shiro: ST2 Inoshishi Aka: ST1 Hayabusa Shared Team 09:15 POOL B #4 Shiro: SI5 Hana Ito Aka: SI4 Chie Mori Shared Individual 09:15 SEMIFINALS #2 Shiro: ST4 Tanuki Aka: ST3 Kitsune Shared Team 09:20 POOL B #5 Shiro: SI6 Lea Roth Aka: SI4 Chie Mori Shared Individual 09:25 POOL B #6 Shiro: SI6 Lea Roth Aka: SI5 Hana Ito Shared Individual SHIAIJO B 6 matches 09:00 POOL C Next up Shiro: SI8 Finn Hale Aka: SI7 Ben Carter","tvShot":"scripts/e2e/output/shared-shiaijo/07-s7-tv-A-before-play.png","scheduleShot":"scripts/e2e/output/shared-shiaijo/08-s7-public-schedule-before-play.png"}
S2-S9-S3	S2 advance	Finish + Start Next →	V2 doubleTap	`{"gapMs":29,"withinWindow":true,"committedByDouble":false,"stillArmed":true,"shot":"scripts/e2e/output/shared-shiaijo/09-s2-finish-double-tapped.png"}`
S2-S9-S3	S2 advance	Finish + Start Next (confirmed)	correct path	`{"next":{"comp":"Shared Individual","eyebrow":"Shared Individual · Pool A · Match 2 of 3","court":"A","shiro":"Jun Oda","aka":"Aoi Sato"},"stUpNextAt":"09:12","shot":"scripts/e2e/output/shared-shiaijo/10-s2-next-si-bout.png"}`
S2-S9-S3	S4a correct	Correct an SI bout while SI's next bout runs	correct path (refused, bc-crpn)	{"liveBout":{"shiro":"Jun Oda","aka":"Aoi Sato"},"notice":"Shiaijo A is running Pool A · Match 2 · Jun Oda vs Aoi Sato. Finish it or send it back to the queue first. Then correct this match.","toast":"⚠️\nShiaijo A is running Pool A · Match 2 · Jun Oda vs Aoi Sato. Finish it or send it back to the queue first. Then correct this match.\n×","editorStillOnLiveBout":true,"sendBack":{"label":"Send back to queue","prominence":"primary","otherLabels":["Cancel"],"title":"Send back to queue?","message":"Jun Oda vs Aoi Sato will be returned to the upcoming queue. Any score entered is kept: starting it again carries on from there."},"shot":"scripts/e2e/output/shared-shiaijo/11-s4a-correction-open.png"}
S2-S9-S3	S4a correct	Back to court	V5 lost place	`{"officiating":"Shared Individual","upNext":{"head":"09:05 · Shared Individual · Match 2 of 3","names":["Jun Oda","Aoi Sato"]},"editor":null,"idle":true,"shot":"scripts/e2e/output/shared-shiaijo/12-s4a-back-to-court.png"}`
S2-S9-S3	S2b advance	Kiken - Voluntary (Shiro), side picked by the operator	V4 hasty (inline prompt, bc-dsid)	`{"preselected":null,"recordDisabledBeforePick":true,"recordLabel":"Record: SHIRO withdrew","chosen":"shiro","withdrawn":"Jun Oda","next":{"comp":"Shared Individual","eyebrow":"Shared Individual · Pool B · Match 1 of 3","court":"A","shiro":"Hana Ito","aka":"Chie Mori"},"nextSiAt":"09:15","stEarliestAt":"09:12","barredRow":"Jun Oda withdrew: record the fusensho. Record fusensho for Eri Kato","shot":"scripts/e2e/output/shared-shiaijo/13-s2b-advance-after-kiken.png"}`
S2-S9-S3	S2b advance	clock check	note	`{"stEarlierThanNextSi":true}`
S2-S9-S3	S7 public	TV board A and the public schedule while SI runs on A	P8 record (D2)	{"consoleRunsNext":"the next SI bout on A (the console stays on SI)","tv":"Shared Court Cup · SHIAIJO A Shared Individual POOL B · 0 / 3 MATCHES SI5 Hana Ito vs Chie Mori SI4 SI6 Lea Roth vs Chie Mori SI4 SI6 Lea Roth vs Hana Ito SI5 NEXT ST2Inoshishi vs HayabusaST1","schedule":"← SHARED COURT CUP Schedule All matches across courts and competitions Filter by player, team, dojo or number… All competitions Shared Individual Shared Team All Shiaijo Shiaijo A Shiaijo B 14 matches of 14 12 Oct 2026 SHIAIJO A 8 matches · in progress ● NOW 09:15 POOL B Shiro: SI5 Hana Ito Aka: SI4 Chie Mori Shared Individual 09:10 POOL A Shiro: SI3 Jun Oda Withdrawn Aka: SI2 Eri Kato Shared Individual 09:12 SEMIFINALS Next up Shiro: ST2 Inoshishi Aka: ST1 Hayabusa Shared Team 09:15 SEMIFINALS #2 Shiro: ST4 Tanuki Aka: ST3 Kitsune Shared Team 09:20 POOL B Next up Shiro: SI6 Lea Roth Aka: SI4 Chie Mori Shared Individual 09:25 POOL B #2 Shiro: SI6 Lea Roth Aka: SI5 Hana Ito Shared Individual 09:00 POOL A Shiro: SI2 Eri Kato Aka: SI1 Aoi Sato Shared Individual – vs M 09:05 POOL A Shiro: SI3 Jun Oda Aka: SI1 Aoi Sato Shared Individual Kiken vs ○○ SHIAIJO B 6 matches 09:00 POOL C Next up Shi","tvShot":"scripts/e2e/output/shared-shiaijo/14-s7-tv-A-while-si-runs.png","scheduleShot":"scripts/e2e/output/shared-shiaijo/15-s7-public-schedule-while-si-runs.png"}
S2-S9-S3	S9 scores page	Finish + Start Next →	correct path	{"listOnA":["09:15 Pool B · Match 1 Shared Individual A ▾ Shiro: SI5 Hana Ito Thames vs Aka: Chie Mori SI4 Musashi Score","09:10 Pool A · Match 3 Shared Individual A ▾ Shiro: SI3 Jun Oda Seishinkan vs Aka: Eri Kato SI2 Mumeishi Score Jun Oda withdrew: record the fusensho. Record fusensho for Eri Kato","09:12 Match 1 Shared Team A ▾ Shiro: ST2 Inoshishi Minami Dojo vs Aka: Hayabusa ST1 Kita Dojo Score Lineup","09:15 Match 2 Shared Team A ▾ Shiro: ST4 Tanuki Nishi Dojo vs Aka: Kitsune ST3 Higashi Dojo Score Lineup","09:20 Pool B · Match 2 Shared Individual A ▾ Shiro: SI6 Lea Roth Hizen vs Aka: Chie Mori SI4 Musashi Score","09:25 Pool B · Match 3 Shared Individual A ▾ Shiro: SI6 Lea Roth Hizen vs Aka: Hana Ito SI5 Thames Score","09:00 Pool A · Match 1 Shared Individual A ▾ Shiro: SI2 Eri Kato Mumeishi – vs M Aka: Aoi Sato SI1 Gyokusen Correct","09:05 Pool A · Match 2 Shared Individual A ▾ Shiro: SI3 Jun Oda Seishinkan Kiken vs ○○ Aka: Aoi Sato SI1 Gyokusen Correct"],"next":{"eyebrow":"SHARED INDIVIDUAL · POOL B · MATCH 2 OF 3","title":"Shiaijo A · 09:20","names":["Lea Roth","Chie Mori"],"court":"A","key":"SHARED INDIVIDUAL · POOL B · MATCH 2 OF 3 \| Shiaijo A · 09:20 \| Lea Roth \| Chie Mori","shiro":"Lea Roth","aka":"Chie Mori"},"expected":{"time":"09:20","label":"Pool B · Match 2","court":"A","shiro":"Lea Roth","aka":"Chie Mori","status":"scheduled","button":"Score"},"shot":"scripts/e2e/output/shared-shiaijo/16-s9-scores-page-finish-next.png"}
S2-S9-S3	S9 scores page	Next x2, Prev x2	correct path (allowed to walk the court)	`{"walked":["SHARED INDIVIDUAL · POOL A · MATCH 3 OF 3","SHARED TEAM · SEMIFINALS · MATCH 1","SHARED INDIVIDUAL · POOL A · MATCH 3 OF 3","SHARED INDIVIDUAL · POOL B · MATCH 2 OF 3"],"comps":["Shared Individual","Shared Team","Shared Individual","Shared Individual"],"shot":"scripts/e2e/output/shared-shiaijo/17-s9-prev-next-walked.png"}`
S2-S9-S3	S9 scores page	Fusenpai (Aka) in the overlay	correct path (after-decision advance)	`{"preselected":null,"recordDisabledBeforePick":true,"recordLabel":"Record: AKA did not show up","chosen":"aka","next":{"eyebrow":"SHARED INDIVIDUAL · POOL B · MATCH 2 OF 3","title":"Shiaijo A · 09:20","names":["Lea Roth","Chie Mori"],"court":"A","key":"SHARED INDIVIDUAL · POOL B · MATCH 2 OF 3 \| Shiaijo A · 09:20 \| Lea Roth \| Chie Mori","shiro":"Lea Roth","aka":"Hana Ito"},"shot":"scripts/e2e/output/shared-shiaijo/18-s9-after-fusenpai.png"}`
S2-S9-S3	S9 scores page	SI's last startable bout on A	correct path	`{"finishLabel":"Finish"}`
S2-S9-S3	S3 switch	picker -> Shared Team while an SI bout runs with a point	correct path	`{"editor":{"eyebrow":"Shared Individual · Pool B · Match 3 of 3","court":"A","title":"Shiaijo A · 09:25 ● Synced","pill":null,"names":["Lea Roth","Hana Ito"]},"upNext":{"head":"09:12 · Shared Team · Match 1","names":["Inoshishi","Hayabusa"]},"nudges":[{"text":"1 Shared Individual match also waiting on this court.","quiet":true}],"shot":"scripts/e2e/output/shared-shiaijo/19-s3-picked-st-si-running.png"}`
S2-S9-S3	S3 switch	reload with Shared Team picked and an SI bout running	V3 interrupt reload (P1)	`{"officiatingAfter":"Shared Individual","shot":"scripts/e2e/output/shared-shiaijo/20-s3-reload-after-pick.png"}`
S2-S9-S3	S3 switch	Start match (ST's Up next)	V1 tapNeighbour right	`{"hit":"Enter lineup","lineupPanelOpened":true,"stillRunning":{"comp":"Shared Individual","eyebrow":"Shared Individual · Pool B · Match 3 of 3","court":"A","shiro":"Lea Roth","aka":"Hana Ito"},"shot":"scripts/e2e/output/shared-shiaijo/21-s3-start-neighbour-tapped.png"}`
S2-S9-S3	S3 switch	Start match (ST's Up next) while the SI bout runs	V2 doubleTap	{"gapMs":27,"withinWindow":true,"writes":["POST Pool B-2/revert-to-queue -> 200 (+100ms)","PUT m-r1-0/score -> 200 (+102ms)","POST Pool B-2/revert-to-queue -> 200 (+188ms)","PUT m-r1-0/score -> 200 (+196ms)"],"runningOnA":[{"court":"A","comp":"Shared Team","label":"Match 1","shiro":"Inoshishi","aka":"Hayabusa","status":"running"}],"alsoRunning":null,"nudges":[{"text":"2 Shared Individual matches also waiting on this court.","quiet":true}],"shot":"scripts/e2e/output/shared-shiaijo/22-s3-st-started.png"}
S2-S9-S3	S3 switch	Finish + Start Next on Shared Team's semi-final	correct path	`{"played":{"shiro":"Inoshishi","aka":"Hayabusa"},"next":{"comp":"Shared Team","eyebrow":"Shared Team · Semifinals · Match 2","court":"A","shiro":"Tanuki","aka":"Kitsune"},"shot":"scripts/e2e/output/shared-shiaijo/23-s3-st-finish-next.png"}`
S2-S9-S3	S3 switch	Start the sent-back SI bout (ST semi-final running)	correct path (bc-sbq)	`{"runningOnA":[{"court":"A","comp":"Shared Individual","label":"Pool B · Match 3","shiro":"Lea Roth","aka":"Hana Ito","status":"running"}],"marksBefore":["M"],"marksAfter":["M"],"shot":"scripts/e2e/output/shared-shiaijo/24-s3-si-bout-back-with-point.png"}`
S2-S9-S3	S2b advance	Fusenpai on SI's last startable bout on A	correct path (nothing to start)	{"preselected":null,"recordDisabledBeforePick":true,"recordLabel":"Record: AKA did not show up","chosen":"aka","runningOnA":[],"court":{"officiating":"Shared Individual","pickerOptions":["Shared Individual","Shared Team"],"upNext":null,"upcoming":[{"head":"09:10 · Shared Individual · Match 3 of 3","names":["Jun Oda","Eri Kato"]}],"later":[{"head":"09:40 · Shared Individual · Match 5","names":["Winner of M2","Winner of M1"]},{"head":"09:45 · Shared Individual · Match 7","names":["Winner of M6","Winner of M5"]}],"completed":5,"editor":null,"nudges":[{"text":"1 Shared Team match also waiting on this court.","quiet":true}],"alsoRunning":null,"done":null,"idle":true},"shot":"scripts/e2e/output/shared-shiaijo/25-s2b-nothing-left-to-start.png"}
S4b-S5	S4b correct	Shared Team semi-final started again	correct path (bc-sbq)	`{"bout1Shiro":["M"]}`
S4b-S5	S4b correct	picker -> Shared Individual to reach its completed bouts	P4 record	`{"completedWhileST":1,"completedWhileSI":5,"shot":"scripts/e2e/output/shared-shiaijo/26-s4b-si-completed-while-st-runs.png"}`
S4b-S5	S4b correct	Correct the SI kiken while Shared Team runs	correct path (refused, bc-crpn; P2 copy)	{"liveBout":{"eyebrow":"Shared Team · Semifinals · Match 2","court":"A","shiro":"Tanuki","aka":"Kitsune"},"notice":"Shiaijo A is running Match 2 · Tanuki vs Kitsune. Finish it or send it back to the queue first. Then correct this match.","toast":"⚠️\nShiaijo A is running Match 2 · Tanuki vs Kitsune. Finish it or send it back to the queue first. Then correct this match.\n×","sendBack":{"label":"Send back to queue","prominence":"primary","otherLabels":["Cancel"],"title":"Send back to queue?","message":"Tanuki vs Kitsune will be returned to the upcoming queue. Any score entered is kept: starting it again carries on from there."},"namesCompetition":false,"shot":"scripts/e2e/output/shared-shiaijo/27-s4b-correction-open.png"}
S4b-S5	S6d reopen	head table starts Shared Team's semi-final while the console holds an SI correction	two devices	`{"alert":"Another bout is running on Shiaijo A Tanuki vs Kitsune · Shared Team Save or cancel the correction (Back to court) to score it here.","shot":"scripts/e2e/output/shared-shiaijo/28-s6d-also-running-during-correction.png"}`
S4b-S5	S6d reopen	Clear kiken and reopen while ST runs	V4 loudest button in the court_busy panel	{"conflictHead":"Shiaijo A is running Match 2 · Tanuki vs Kitsune. Finish it or send it back to the queue first.","namesCompetition":false,"loud":"Queue it and reopen","others":["Leave it running"],"runningOnA":[{"court":"A","comp":"Shared Individual","label":"Pool A · Match 2","shiro":"Jun Oda","aka":"Aoi Sato","status":"running"}],"stSemi":{"time":"09:15","label":"Match 2","court":"A","shiro":"Tanuki","aka":"Kitsune","status":"scheduled","button":"Score"},"conflictShot":"scripts/e2e/output/shared-shiaijo/29-s6d-reopen-court-busy-panel.png","shot":"scripts/e2e/output/shared-shiaijo/30-s6d-after-queue-it-and-reopen.png"}
S4b-S5	S6d reopen	reload right after the reopen	V3/V5 interrupt reload (P1, P7)	{"court":{"officiating":"Shared Individual","pickerOptions":["Shared Individual","Shared Team"],"upNext":{"head":"09:10 · Shared Individual · Match 3 of 3","names":["Jun Oda","Eri Kato"]},"upcoming":[],"later":[{"head":"09:40 · Shared Individual · Match 5","names":["Winner of M2","Winner of M1"]},{"head":"09:45 · Shared Individual · Match 7","names":["Winner of M6","Winner of M5"]}],"completed":4,"editor":{"eyebrow":"Shared Individual · Pool A · Match 2 of 3","court":"A","title":"Shiaijo A · 09:05 ● Synced","pill":null,"names":["Jun Oda","Aoi Sato"]},"nudges":[{"text":"1 Shared Team match also waiting on this court.","quiet":true}],"alsoRunning":null,"done":null,"idle":false},"shot":"scripts/e2e/output/shared-shiaijo/31-s6d-reload.png"}
S4b-S5	S5 SI done on A	play the rest of SI on A	correct path	`{"played":[{"shiro":"Jun Oda","aka":"Aoi Sato"},{"shiro":"Jun Oda","aka":"Eri Kato"}]}`
S4b-S5	S5 SI done on A	the console with SI pools done on A, B pools still to play	P3 / P7 record	{"court":{"officiating":"Shared Individual","pickerOptions":["Shared Individual","Shared Team"],"upNext":null,"upcoming":[],"later":[{"head":"09:40 · Shared Individual · Match 5","names":["Winner of M2","Winner of M1"]},{"head":"09:45 · Shared Individual · Match 7","names":["Winner of M6","Winner of M5"]}],"completed":6,"editor":null,"nudges":[{"text":"Switch to Shared Team: 1 match waiting on this court.","quiet":false}],"alsoRunning":null,"done":null,"idle":true},"nudges":[{"text":"Switch to Shared Team: 1 match waiting on this court.","quiet":false,"amber":true}],"shot":"scripts/e2e/output/shared-shiaijo/32-s5-ii-si-complete-on-A.png"}
S4b-S5	S5 SI done on A	reload (the explicit pick is dropped)	V3 interrupt reload (P1)	`{"officiatingAfter":"Shared Team","shot":"scripts/e2e/output/shared-shiaijo/33-s5-ii-after-reload.png"}`
S4b-S5	S5 SI knockout on A	court B finishes its pools; court A, untouched	correct path (P7: held, no reload)	{"bPlayed":6,"court":{"officiating":"Shared Individual","pickerOptions":["Shared Individual","Shared Team"],"upNext":{"head":"09:30 · Shared Individual · Match 1","names":["Finn Hale","Aoi Sato"]},"upcoming":[{"head":"09:35 · Shared Individual · Match 2","names":["Goro Abe","Lea Roth"]}],"later":[{"head":"09:40 · Shared Individual · Match 5","names":["Winner of M2","Winner of M1"]},{"head":"09:45 · Shared Individual · Match 7","names":["Winner of M6","Winner of M5"]}],"completed":6,"editor":null,"nudges":[{"text":"1 Shared Team match also waiting on this court.","quiet":true}],"alsoRunning":null,"done":null,"idle":true},"shot":"scripts/e2e/output/shared-shiaijo/34-s5-quarter-finals-arrive-on-A.png"}
S4b-S5	S5 SI knockout on A	quarter-finals on both courts	correct path	`{"quarters":[{"court":"B","played":{"shiro":"Eri Kato","aka":"Ben Carter","winner":"aka"}},{"court":"B","resolved":"Record fusensho for Dan Evans"},{"court":"A","played":{"shiro":"Finn Hale","aka":"Aoi Sato","winner":"aka"}},{"court":"A","played":{"shiro":"Goro Abe","aka":"Lea Roth","winner":"shiro"}}]}`
S4b-S5	S5 SI knockout on A	Finish on SI semi-final M5	V2 retapWhileSaving	`{"pair":{"shiro":"Goro Abe","aka":"Aoi Sato"},"savingSeen":true,"retapped":true,"labelAfter":null}`
S4b-S5	S5 SI final waits	the console with SI waiting on a feeder from B	P3 / P7 record	{"court":{"officiating":"Shared Individual","pickerOptions":["Shared Individual","Shared Team"],"upNext":null,"upcoming":[],"later":[{"head":"09:45 · Shared Individual · Match 7","names":["Winner of M6","Aoi Sato"]}],"completed":8,"editor":null,"nudges":[{"text":"Switch to Shared Team: 1 match waiting on this court.","quiet":false}],"alsoRunning":null,"done":null,"idle":true},"nudges":[{"text":"Switch to Shared Team: 1 match waiting on this court.","quiet":false,"amber":true}],"shot":"scripts/e2e/output/shared-shiaijo/35-s5-i-final-waits-on-B.png"}
S4b-S5	S5 SI final waits	tap the nudge line (Shared Team)	correct path	`{"officiating":"Shared Team","upNext":{"head":"09:15 · Shared Team · Match 2","names":["Tanuki","Kitsune"]},"shot":"scripts/e2e/output/shared-shiaijo/36-s5-switched-by-tap.png"}`
S6a	S6a move	schedule page: move running SI M6 from B onto A while ST runs on A	P5 record	{"askedFirst":false,"runningOnA":[{"court":"A","comp":"Shared Team","label":"Match 2","shiro":"Tanuki","aka":"Kitsune","status":"running"},{"court":"A","comp":"Shared Individual","label":"Match 6","shiro":"Dan Evans","aka":"Ben Carter","status":"running"}],"alert":"Another bout is running on Shiaijo A Dan Evans vs Ben Carter · Shared Individual A court runs one bout at a time. Score or correct these from their competition's admin view.","consoleShot":"scripts/e2e/output/shared-shiaijo/37-s6a-two-running-on-A.png","scheduleShot":"scripts/e2e/output/shared-shiaijo/38-s6a-schedule-after-move.png"}
S6a	S6a move	a point on the bout the console shows, two bouts running	P5 record	`{"held":{"comp":"Shared Team","eyebrow":"Shared Team · Semifinals · Match 2","court":"A","shiro":"Tanuki","aka":"Kitsune"},"pointTried":{"comp":"Shared Team","before":0,"after":1},"sync":"Synced","messages":"⚠️ Shiaijo A is running Match 6 (Semifinals). Finish it or send it back to the queue first. ×","shot":"scripts/e2e/output/shared-shiaijo/39-s6a-point-with-two-running.png"}`
S6a	S6a move	Send back to queue (the shown bout)	V4 hastyShiaijoConfirm (recovery)	{"label":"Send back to queue","prominence":"primary","otherLabels":["Cancel"],"title":"Send back to queue?","message":"Tanuki vs Kitsune will be returned to the upcoming queue. Any score entered is kept: starting it again carries on from there.","runningOnA":[{"court":"A","comp":"Shared Individual","label":"Match 6","shiro":"Dan Evans","aka":"Ben Carter","status":"running"}],"court":{"officiating":"Shared Team","pickerOptions":["Shared Individual","Shared Team"],"upNext":{"head":"09:15 · Shared Team · Match 2","names":["Tanuki","Kitsune"]},"upcoming":[],"later":[{"head":"10:02 · Shared Team · Match 3","names":["Winner of M2","Inoshishi"]}],"completed":1,"editor":{"eyebrow":"Shared Individual · Semifinals · Match 6","court":"A","title":"Shiaijo A · 09:40 ● Synced","pill":null,"names":["Dan Evans","Ben Carter"]},"nudges":[],"alsoRunning":null,"done":null,"idle":false},"shot":"scripts/e2e/output/shared-shiaijo/40-s6a-after-send-back.png"}
S6a	S6a move	the refused point, read back on the head table after the send back	P5 record	`{"bout2ShiroOnBoardBefore":1,"bout2ShiroStored":0,"shot":"scripts/e2e/output/shared-shiaijo/41-s6a-refused-point-read-back.png"}`

gitrgoliveira · 2026-10-12T09:33:05Z

Audit rows, current run: Moved match (bc-mvsj) and smoke

Raw rows from make e2e on d3894b4 (2026-10-12, main merged through #467), read from each journey's audit.json. One row per operator action x clumsy variant; the Observed column is what the journey recorded, with screenshot paths under scripts/e2e/output/. These rows carry no judgement: the 2026-09-24 audit comments above are the review as performed, with the reviewers' judgement.

Journey	Step	Action	Variant	Observed
MV	setup	wizard defaults, both competitions created before any roster	seed	`{"defaults":{"IND":{"startTime":"09:00","sharedCourtHint":"Concurrency for this competition equals the number of shiaijo (courts) assigned. Different competitions can share shiaijo (courts); the schedule prevents conflicts."},"TEAMS":{"startTime":"09:30","sharedCourtHint":"Concurrency for this competition equals the number of shiaijo (courts) assigned. Different competitions can share shiaijo (courts); the schedule prevents conflicts."}}}`
MV	setup	draw read back (Scores tabs)	seed	{"individual":[{"time":"09:00","label":"League table · Match 1","court":"A","shiro":"Pia Lund","aka":"Oto Kudo","status":"scheduled","button":"Score"},{"time":"09:05","label":"League table · Match 2","court":"A","shiro":"Rin Saito","aka":"Nils Berg","status":"scheduled","button":"Score"},{"time":"09:10","label":"League table · Match 3","court":"A","shiro":"Oto Kudo","aka":"Mika Endo","status":"scheduled","button":"Score"},{"time":"09:15","label":"League table · Match 4","court":"A","shiro":"Pia Lund","aka":"Rin Saito","status":"scheduled","button":"Score"},{"time":"09:20","label":"League table · Match 5","court":"A","shiro":"Nils Berg","aka":"Mika Endo","status":"scheduled","button":"Score"},{"time":"09:25","label":"League table · Match 6","court":"A","shiro":"Rin Saito","aka":"Oto Kudo","status":"scheduled","button":"Score"},{"time":"09:30","label":"League table · Match 7","court":"A","shiro":"Pia Lund","aka":"Nils Berg","status":"scheduled","button":"Score"},{"time":"09:35","label":"League table · Match 8","court":"A","shiro":"Rin Saito","aka":"Mika Endo","status":"scheduled","button":"Score"},{"time":"09:40","label":"League table · Match 9","court":"A","shiro":"Nils Berg","aka":"Oto Kudo","status":"scheduled","button":"Score"},{"time":"09:45","label":"League table · Match 10","court":"A","shiro":"Pia Lund","aka":"Mika Endo","status":"scheduled","button":"Score"}],"teams":[{"time":"09:30","label":"Match 1","court":"B","shiro":"Inoshishi","aka":"Hayabusa","status":"scheduled","button":"Score"},{"time":"09:46","label":"Match 2","court":"B","shiro":"Tanuki","aka":"Kitsune","status":"scheduled","button":"Score"}]}
MV	1 setup	a bout running on each court	correct path	`{"courtA":{"shiro":"Pia Lund","aka":"Oto Kudo"},"courtB":{"shiro":"Inoshishi","aka":"Hayabusa"},"shotA":"scripts/e2e/output/moved-match/01-court-A-bout-running.png","shotB":"scripts/e2e/output/moved-match/02-court-B-teams-running.png"}`
MV	2 court B before	officiating title on court B	correct path	`{"title":"Teams","picker":false,"state":{"officiating":"Teams","pickerOptions":null,"upNext":{"head":"09:46 · Teams · Match 2","names":["Tanuki","Kitsune"]},"upcoming":[],"later":[{"head":"10:02 · Teams · Match 3","names":["Winner of M2","Winner of M1"]}],"completed":0,"editor":{"eyebrow":"Teams · Semifinals · Match 1","court":"B","title":"Shiaijo B · 09:30 ● Synced","pill":null,"names":["Inoshishi","Hayabusa"]},"nudges":[],"alsoRunning":null,"done":null,"idle":false}}`
MV	3 move	court control on the queue row	size	`{"w":30,"h":44,"meets44":false}`
MV	3 move	Move to Shiaijo B confirm	V4 hastyShiaijoConfirm	`{"label":"Move to Shiaijo B","prominence":"primary","otherLabels":["Cancel"],"title":"Move to Shiaijo B?","message":"Rin Saito vs Mika Endo leaves Shiaijo A and joins the queue on Shiaijo B.","moved":{"shiro":"Rin Saito","aka":"Mika Endo"},"confirmShot":"scripts/e2e/output/moved-match/03-move-confirm.png"}`
MV	3 move	court A's queue after the move	correct path	`{"queue":["Rin Saito vs Nils Berg","Oto Kudo vs Mika Endo","Pia Lund vs Rin Saito","Nils Berg vs Mika Endo","Rin Saito vs Oto Kudo","Pia Lund vs Nils Berg","Nils Berg vs Oto Kudo","Pia Lund vs Mika Endo"],"shot":"scripts/e2e/output/moved-match/04-court-A-after-move.png"}`
MV	4 court B live	court B after the move (no reload)	correct path	`{"options":["Individual","Teams"],"officiating":"Teams","queue":["Tanuki vs Kitsune"],"nudges":[{"text":"1 Individual match also waiting on this court.","quiet":true,"amber":false}],"shot":"scripts/e2e/output/moved-match/05-court-B-picker-appeared.png"}`
MV	5 pick Individual	competition picker -> Individual	correct path	`{"upNext":{"shiro":"Rin Saito","aka":"Mika Endo"},"editorStill":{"eyebrow":"Teams · Semifinals · Match 1","court":"B","title":"Shiaijo B · 09:30 ● Synced","pill":null,"names":["Inoshishi","Hayabusa"]},"shot":"scripts/e2e/output/moved-match/06-court-B-individual-picked.png"}`
MV	5a start refused	Start match on the moved match from the Scores tab while Teams runs on B	correct path (court gate)	`{"refusal":"⚠️ Shiaijo B is running Match 1 (Semifinals). Finish it or send it back to the queue first. ×","runningBout":"Teams · Semifinals · Match 1","namesCompetition":false,"namesTeams":false,"shot":"scripts/e2e/output/moved-match/07-scores-tab-start-refused.png"}`
MV	5b console start	Start match (Up next) while the Teams bout is fought	correct path (bc-sbq: the running bout is sent back)	`{"runningOnB":[{"court":"B","comp":"Individual","label":"League table · Match 8","shiro":"Rin Saito","aka":"Mika Endo","status":"running"}],"alsoRunning":null,"nudges":[{"text":"2 Teams matches also waiting on this court.","quiet":true}],"shot":"scripts/e2e/output/moved-match/08-court-B-moved-match-started.png"}`
MV	5b console start	Teams semi-final started again	recovery	`{"pointKept":true,"shot":"scripts/e2e/output/moved-match/09-court-B-teams-restarted-with-point.png"}`
MV	6 finish elsewhere	head table finishes the Teams semi-final	two devices	{"headFinishLabel":"Finish + Start Next →","courtB":{"officiating":"Teams","pickerOptions":["Individual","Teams"],"upNext":null,"upcoming":[],"later":[{"head":"10:02 · Teams · Match 3","names":["Winner of M2","Inoshishi"]}],"completed":1,"editor":{"eyebrow":"Teams · Semifinals · Match 2","court":"B","title":"Shiaijo B · 09:46 ● Synced","pill":null,"names":["Tanuki","Kitsune"]},"nudges":[{"text":"1 Individual match also waiting on this court.","quiet":true}],"alsoRunning":null,"done":null,"idle":false},"shot":"scripts/e2e/output/moved-match/10-court-B-after-head-table-finish.png"}
MV	6 finish elsewhere	court B plays the other semi-final	correct path	`{"pair":{"shiro":"Tanuki","aka":"Kitsune"}}`
MV	6 nudge	court B once Teams has nothing left, 3s later	correct path	{"courtB":{"officiating":"Teams","pickerOptions":["Individual","Teams"],"upNext":null,"upcoming":[],"later":[],"completed":3,"editor":null,"nudges":[{"text":"Switch to Individual: 1 match waiting on this court.","quiet":false}],"alsoRunning":null,"done":"Teams is complete on Shiaijo B3 matches scored. Another competition still has matches on this court: switch above.","idle":false},"nudges":[{"text":"Switch to Individual: 1 match waiting on this court.","quiet":false,"amber":true}],"shot":"scripts/e2e/output/moved-match/11-court-B-switch-to-individual.png"}
MV	6 nudge	tap the Switch line	correct path	`{"officiating":"Individual","upNext":{"shiro":"Rin Saito","aka":"Mika Endo"},"shot":"scripts/e2e/output/moved-match/12-court-B-switched-by-tap.png"}`
MV	7 other door	head table moves a waiting match A -> B from the Scores tab	correct path	`{"moved":{"shiro":"Pia Lund","aka":"Mika Endo"},"confirmAsked":false,"courtA":["Rin Saito vs Nils Berg","Oto Kudo vs Mika Endo","Pia Lund vs Rin Saito","Nils Berg vs Mika Endo","Rin Saito vs Oto Kudo","Pia Lund vs Nils Berg","Nils Berg vs Oto Kudo"],"courtB":["Rin Saito vs Mika Endo","Pia Lund vs Mika Endo"],"shotA":"scripts/e2e/output/moved-match/13-court-A-after-scores-tab-move.png","shotB":"scripts/e2e/output/moved-match/14-court-B-after-scores-tab-move.png"}`
MV	8 second start	Start match on Pia Lund vs Mika Endo at once after starting Rin Saito vs Mika Endo	V2 hurried second start (inside the list refresh)	{"notice":"Shiaijo B is running League table · Match 8. Finish it or send it back to the queue first.","courtB":{"officiating":"Individual","pickerOptions":["Individual","Teams"],"upNext":{"head":"09:45 · Individual · Match 10 of 10","names":["Pia Lund","Mika Endo"]},"upcoming":[],"later":[],"completed":0,"editor":{"eyebrow":"Individual · League table · Match 8 of 10","court":"B","title":"Shiaijo B · 09:35 ● Synced","pill":null,"names":["Rin Saito","Mika Endo"]},"nudges":[],"alsoRunning":null,"done":null,"idle":false},"runningOnB":[{"court":"B","comp":"Individual","label":"League table · Match 8","shiro":"Rin Saito","aka":"Mika Endo","status":"running"}],"shot":"scripts/e2e/output/moved-match/15-court-B-second-start-at-once.png"}
MV	8 competitor busy	Start match on Pia Lund vs Mika Endo (one of them fights on A) while Rin Saito vs Mika Endo runs on B	V5 lost place (the moved match's competitor is elsewhere)	{"notice":"Pia Lund is fighting now in League table · Match 1 on Shiaijo A. Finish that match first.","courtB":{"officiating":"Individual","pickerOptions":["Individual","Teams"],"upNext":{"head":"09:35 · Individual · Match 8 of 10","names":["Rin Saito","Mika Endo"]},"upcoming":[{"head":"09:45 · Individual · Match 10 of 10","names":["Pia Lund","Mika Endo"]}],"later":[],"completed":0,"editor":{"eyebrow":"Individual · League table · Match 8 of 10","court":"B","title":"Shiaijo B · 09:35","pill":"PRE-MATCH","names":["Rin Saito","Mika Endo"]},"nudges":[],"alsoRunning":null,"done":null,"idle":false},"runningOnB":[],"shot":"scripts/e2e/output/moved-match/16-court-B-start-refused-competitor-busy.png"}
		Discard draw	V4 hastyConfirm	`{"label":"Discard draw","prominence":"danger","otherLabels":["Cancel"],"title":"Please confirm","message":"Discard the generated draw for \"Men Individual\"? The pools/bracket will be removed so you can edit participants and settings, then generate a new draw.","drawDiscarded":true}`
		Call to court	V2 doubleTap	`{"gapMs":16,"withinWindow":true,"labelAfter":"Call again"}`
		running bout	V3 interrupt	`{"kind":"reload","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","editorBack":true,"sameMatch":true}`
		bout with one ippon	V3 interrupt	`{"kind":"hidden","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","resumed":true,"method":"document.hidden override + dispatched visibilitychange","sameMatch":true}`
		bout with one ippon	V3 interrupt	`{"kind":"back","before":"/admin/shiaijo/A","landedOn":"/admin/competition/men-individual/scores","hadHistory":true,"after":"/admin/shiaijo/A","editorBack":true,"sameMatch":true,"shiroPointsKept":1}`
		Shiro M	V1 tapNeighbour	`{"intended":{"tag":"button","label":"M","testid":null,"className":"ipt-btn ","disabled":false},"hit":{"tag":"button","label":"K","testid":null,"className":"ipt-btn ","disabled":false},"gapPx":4}`
		Aka M tapped offline	V3 interrupt	`{"kind":"offline","before":"/admin/shiaijo/A","after":"/admin/shiaijo/A","stateWhileOffline":"Syncing…","resynced":true}`
		Finish (armed)	V2 retapWhileSaving	`{"savingSeen":true,"retapped":true,"labelAfter":null,"completedIsBout2":true,"result":"K X M"}`

A fresh-clone run lost J1's reset tap: removing Aka's M and tapping M again while the removal's write was out put back the value the server still showed, so the board read clean and the removal's SSE echo wiped the new M. J1's reset steps now wait for Synced first; they reset the bout and are not the place to test that race. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

On the individual score editor, taking a mark off and tapping it back while the removal's autosave was out returned the board to the value the server still showed. The editor judged unsaved work by comparing the board with the server, so it read the board as untouched and adopted the removal's SSE echo over the newer tap; the tap's own save named no change. The reverse (add, then remove quickly) brought the mark back. The editor now records when the operator last edited the board (points, fouls, overtime, draw) in the server's clock frame and never adopts a server copy stamped before it, the rule the team sheet's rows already follow (rowTakenFromServer, bc-kclr). Adopting the server's copy clears the stamp. Representative-player picks are not board edits and keep their own adopt. Render test pins both orders and that a later change made elsewhere is still followed (fails without the fix); the e2e test holds the save 150 ms and checks the mark on the board and after a reload (fails 2 of 2 without the fix). One changed-groups fixture now stamps its server copies, as the server does. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Code review of PR #442 found checks that passed without testing anything: - bc-tmfn tapped Finish twice inside the bounce window with errors swallowed; it now asserts the unscored-bout refusal on each tap. - bc-htcr and J2 checked only that Ht existed or the text changed; they now check Ht is in the corrected side's cell. - The kachinuki viewer check matched a team name shown before the final; it now waits for the final's winner mark. - J8 and J2 never asserted that the court moves on after a fusenpai recorded on the court console; they do now. - J1's Scores-tab Next could pass with no navigation; it now needs two distinct editors, all on court B. - moved-match, S4a and the shared-court "nothing started" checks read after a fixed sleep or compared empty with empty; they now wait for the event, score b2 first, and poll the whole window. - The shared-court double tap asserts the taps were a real double tap. - enterAdmin and enterTournament no longer take the sign-in branch when the first paint is slower than 5 s. - The individual spec's bead guards are no longer serial with the journeys, so a failing journey skips none of them. Docs review: the README no longer says clumsy outcomes are never asserted, lists the prerequisites and what "green" counts, and names both second-device kinds; the allowlist reason names a barred competitor. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

On the court console the first start can land inside the 400 ms bounce window of its tap. The Up next card then shows the next match with its Start match enabled, and the bounce started it and sent the match just started back to the queue (found by PR #442's shared-shiaijo journey in a --workers=1 run: SF1 started, then SF1 sent back and SF2 started). Every Start match on the console (the Up next card and the queue rows) now shares one keyless bounce guard through tap_guard.jsx's acceptTap, as Call to court does (bc-cdbl). A tap after the window and a keyboard activation still start. The render test fails without the fix. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

When the server stored a Finish but the device never processed the answer (a dropped connection, a reload or sleep right after the tap, a reply past the 12 s wait), the queue resent it with the same stamp, as designed. The correction-reason rule read that identical copy as a correction of a finished match and answered 400, so the operator was told a recorded result was rejected (found by J6 in a fresh-clone run). Operator decision, option A: an exact resend of a completed score write (the stored result's own stamp, nothing it names differing from what is stored once read as the write path would land it) is answered as recorded, with the stored match and no correction reason, before any refusal, writing nothing to the match file (engine.ExactScoreReplay, the score counterpart of exactDecisionReplay). A same-stamp write that differs still needs a reason. The client also takes a settled queue entry out of browser storage at once, narrowing the window. CLAUDE.md's correction rule and the queue's replay sentence now agree; openapi and the network architecture page describe the answer. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

… rules Operator rulings 2026-10-12/13: - The separate destructive-ops password is gone. A signed-in organiser is never asked for the password again, except Delete competition, which sends the re-typed password in X-Confirm-Password and is refused (401) unless it matches, checked by the sign-in verifier. - Self-run: completing a competition needs no password; adding and editing participants is open to anyone; organiser setup, corrections and destructive actions need the main password (branding logo and league tie-break routes added to that list). - Self-run mode has no check-in: Competition.CheckInApplies is the one owner the draw, the Swiss field and the estimate ask; the three check-in writes answer 409 check_in_not_in_self_run; the SPA hides the toggle and the registration desk. - A competition with no knockout rounds cannot be completed while a match is unfinished, nor a Swiss one while rounds remain to draw. - TOURNAMENT_ADMIN_PASSWORD_HASH and a stored admin_password are ignored. Review fixes: the exact-replay probe runs the merge on a copy (held groups excluded) and covers kachinuki, engi and fixed-order teams; a self-run participant's exact resend of a landed finish is answered as recorded; the individual and team editors clear an edit stamp once a save built after it lands (shared landedAutosave). Docs state current behaviour throughout. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

# Conflicts: # CLAUDE.md

Operator ruling 2026-10-13: in file mode POST /api/tournament/reset answers only a loopback peer (judged on the TCP peer address, never on forwarded headers); anyone else gets 403 reset_local_only, "Reset the password on the machine running the server.", which the /reset page shows. Locked mode still answers 404. Review fixes: a self-run participant's exact resend of a finish adds no history line; the file-mode password check is constant-time; the self-run setup text names the admin password and says Delete competition asks for it again; the API spec and docs state that anyone can add and edit participants in self-run and when a competition can be completed; gosec annotation on the confirm header name; a useMemo dependency the linter flagged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

A page whose name rebinds to 127.0.0.1, opened in a browser on the server machine, passed the loopback-peer and same-origin checks. The reset now also requires the request's Host to be localhost or a loopback address, else the same 403 reset_local_only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

gitguardian · 2026-10-13T09:22:02Z

️✅ There are no secrets present in this pull request anymore.

If these secrets were true positive and are still valid, we highly recommend you to revoke them.
While these secrets were previously flagged, we no longer have a reference to the
specific commits where they were detected. Once a secret has been leaked into a git
repository, you should consider it compromised, even if it was deleted immediately.
Find here more information about risks.

^{_{🦉 GitGuardian detects secrets in your source code to help developers and security teams secure the modern development process. You are seeing this because you or someone else with access to this repository has authorized GitGuardian to scan your pull request.}}

GitGuardian read the fixture value as a generic password. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

The sign-in dialog, the reset page and confirm, the discarded-writes message, the server's error sentences and startup log, the API spec and the docs said "tournament password" where the setup screens and the operator docs say "admin password". Identifiers, the X-Tournament-Password header and wire codes are unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Infrastructure: retapWhileSaving and doubleTap fail instead of passing silently (the double-tap limit derives from the app's TAP_BOUNCE_MS); the start/reopen 409 allowlist entry matches only a start or a reopen, and console lines pair one-to-one with failed answers; the shared server helper kills its child and removes its data folder on process exit, awaits the child on stop and catches a port race; a failing test gets the server log; afterAll output is attributed to its own file. Journeys: J7 asserts the finished TV board; the team guards no longer hang off one serial chain and the kachinuki guards cannot skip the final; the kachinuki totals, the team kiken and the one-mark double tap are asserted; runningOnCourt waits for the list; J6's second device is its own context; lineup names are read after the source. Docs: the harness README is developer documentation (adding a journey, traces, network needs); the developer guide links to it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Kachinuki guard cleanup is capped and cannot replace a guard's own failure, and each guard failure is annotated when it happens; the overtime guard checks Encho landed before its negative check; dead withinWindow assertions and a vacuous viewer assertion are gone; J7 says what the TV board can show; the shared server helper also cleans up on SIGINT/SIGTERM when nothing else handles them; README states the default timeout, parallel workers, when a server stops and that the results folder is emptied. Docs: the laptop shortcuts sentence states which editors take Left, Right and Esc, as the code does. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

A failed answer is recorded before its body arrives and the console check awaits pending bodies, so an allowed refusal cannot fail a test; wizardDefaults waits for the schedule estimate it expects instead of fixed sleeps; a failing test gets its file's server log even when it never asked for the server; the bc-ltap hold is released after the tap-back rather than after a fixed 150 ms. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

An answer no console line claimed yet is kept for the pairing window, so a refusal whose console line arrives in the next test still pairs; a refusal body is awaited up to the expect timeout (one shared constant) and its timer cleared; consolePolicy is torn down before serverLog, so a test failed by the console policy carries server.log. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

gitrgoliveira and others added 7 commits September 24, 2026 00:49

fix(bc-e2eh): let concurrent e2e runs use separate results folders

2ace9f5

Each playwright run empties outputDir at start, so two runs in one checkout deleted each other's traces (ENOENT on context close). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

test(mp-yqxn.2): assert a coarse pointer before every tap-target meas…

4aeda72

…urement Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

test(mp-yqxn.2): name each journey fixme by its filed bead

b7182b6

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

test(mp-yqxn.2): encode the every-bout-is-fought ruling as bc-tmfn

548a936

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

test(mp-yqxn.2): say team match, not fixed-order (operator ruling)

bae2b33

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Merge remote-tracking branch 'origin/main' into bc-e2eh-operator-review

c366c50

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

gitrgoliveira mentioned this pull request Sep 24, 2026

fix(bc-tmfn): operator fixes: team bouts, withdrawals, knockout rounds and the Bracket page #444

Merged

11 tasks done

gitrgoliveira and others added 2 commits September 25, 2026 20:15

Merge remote-tracking branch 'origin/main' into bc-e2eh-operator-review

4487d89

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

gitrgoliveira mentioned this pull request Oct 2, 2026

fix(bc-rfsw,bc-offl,bc-rboq,bc-flst,bc-kheb,bc-mrgc,bc-hlck): merge match writes by change, held results visible, correction refusals, kachinuki encho #453

Merged

15 tasks done

gitrgoliveira mentioned this pull request Oct 6, 2026

fix: iPad operator fixes; a team's lineup carries between matches #458

Merged

39 tasks done

gitrgoliveira and others added 5 commits October 7, 2026 15:30

Merge branch 'main' into bc-e2eh-operator-review

d2e7d93

Merge remote-tracking branch 'origin/bc-e2eh-operator-review' into bc…

f308676

…-e2eh-operator-review

gitrgoliveira mentioned this pull request Oct 7, 2026

fix(bc-dsid, bc-otpl, bc-htsd, bc-dlck, bc-kosc, bc-crpn): operator review fixes from PR #442 #462

Merged

7 tasks done

Merge remote-tracking branch 'origin/main' into bc-e2eh-operator-review

ad32243

gitrgoliveira and others added 10 commits October 12, 2026 11:17

Merge branch 'main' into bc-e2eh-operator-review

9ef805f

Merge remote-tracking branch 'origin/main' into bc-e2eh-operator-review

3b45f33

# Conflicts: # CLAUDE.md

test(bc-cmpw): use the house test password in the delete dialog test

9d03ba2

GitGuardian read the fixture value as a generic password. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

gitrgoliveira marked this pull request as ready for review October 13, 2026 09:35

Copilot AI balanced review requested due to automatic review settings October 13, 2026 09:35

Copilot AI reviewed Oct 13, 2026

gitrgoliveira and others added 3 commits October 13, 2026 11:48

Copilot AI balanced review requested due to automatic review settings October 13, 2026 11:41

Copilot AI reviewed Oct 13, 2026

gitrgoliveira and others added 2 commits October 13, 2026 13:49

Copilot AI balanced review requested due to automatic review settings October 13, 2026 13:33

Copilot AI reviewed Oct 13, 2026

gitrgoliveira merged commit 723e1fd into main Oct 13, 2026
13 checks passed

gitrgoliveira deleted the bc-e2eh-operator-review branch October 13, 2026 15:29

gitrgoliveira mentioned this pull request Oct 14, 2026

fix(bc-evts): engi tap-back race, one edit-time rule for every score editor; e2e harness re-verified (bc-e2rv) #473

Open

8 tasks done

Navigation Menu

Repository navigation

test(bc-e2eh, mp-yqxn.2): e2e harness and a clumsy-operator review of knockout and mixed - #442

test(bc-e2eh, mp-yqxn.2): e2e harness and a clumsy-operator review of knockout and mixed#442
gitrgoliveira merged 41 commits into
mainfrom
bc-e2eh-operator-review

gitrgoliveira commented Sep 24, 2026 •

edited

Loading

Uh oh!

gitrgoliveira commented Sep 24, 2026 •

edited

Loading

Uh oh!

gitrgoliveira commented Sep 24, 2026 •

edited

Loading

Uh oh!

gitrgoliveira commented Sep 24, 2026 •

edited

Loading

Uh oh!

gitrgoliveira commented Oct 12, 2026

Uh oh!

gitrgoliveira commented Oct 12, 2026

Uh oh!

gitrgoliveira commented Oct 12, 2026

Uh oh!

gitguardian Bot commented Oct 13, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

gitrgoliveira commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Which PR closes which bead

The fix: a tapped-back mark was lost (bc-ltap)

The second fix: a double tap on Start match started the next match (bc-sdtp)

The third fix: a finish whose answer was lost was refused on resend (bc-xrpl)

The fourth change (bc-cmpw): one admin password

Shared shiaijo and the moved match (bc-shcr, bc-mvsj)

Walkthrough: the five-step operator path

1. Lineup entry

2. Navigating bouts and matches

3. Score entry and the running total

4. Special decisions

5. Confirm and advance

What passed

Operator rulings made during this review (2026-09-24)

Findings (severity-ranked)

Harness notes

Files changed

Screenshots

Test plan

Uh oh!

gitrgoliveira commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Persona audit A: individual (J1, J2, J6, J8)

mp-yqxn.2 reviewer A: persona audit, INDIVIDUAL competitions

J1: shiaijo console, pools to knockout, same court (F1, shiaijo A and B)

J2: special decisions (F1b shiaijo E; F2 shiaijo C)

J6: confirm + advance OFFLINE (F2, shiaijo C, bronze match on)

J8: the same scoring from the Pools tab (F1c) and the Bracket tab (F2d)

Uh oh!

gitrgoliveira commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Persona audit B: team, sizes 3 and 5 (J3, J4, J7)

Reviewer B audit: fixed-order TEAM competitions (mp-yqxn.2)

Ruling: three lineup entry points, defect or intentional?

J3: lineup entry from all three entry points (F3 size 3, F4 size 5)

J4: fixed-order team scoring at the court (F4, /admin/shiaijo/D)

J7: one result on every surface (court iPad, head-table iPad, spectator phone, TV board)

Findings (severity-ranked; ids are placeholders for the orchestrator)

Uh oh!

gitrgoliveira commented Sep 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Persona audit C: kachinuki, size 5 (J5)

Reviewer C: J5 kachinuki persona audit (mp-yqxn.2)

Audit table (84 rows)

kachinukiVisiblePositions verdict (admin_scoring_team.jsx:506)

Findings (severity-ranked; file:line re-derived with grep -n on this worktree)

Harness defects and things not reachable through the UI

Uh oh!

gitrgoliveira commented Oct 12, 2026

Audit rows, current run: Kachinuki (J5)

Uh oh!

gitrgoliveira commented Oct 12, 2026

Audit rows, current run: Shared shiaijo (bc-shcr: S1 to S9, P1 to P8)

Uh oh!

gitrgoliveira commented Oct 12, 2026

Audit rows, current run: Moved match (bc-mvsj) and smoke

Uh oh!

gitguardian Bot commented Oct 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

️✅ There are no secrets present in this pull request anymore.

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

gitrgoliveira commented Sep 24, 2026 •

edited

Loading

gitrgoliveira commented Sep 24, 2026 •

edited

Loading

gitrgoliveira commented Sep 24, 2026 •

edited

Loading

gitrgoliveira commented Sep 24, 2026 •

edited

Loading

gitguardian Bot commented Oct 13, 2026 •

edited

Loading