[entelecheia]

Rebased onto main 7ecb50f at fd2e2a5. Only CHANGELOG/contracts context was resolved; all merged #350/#361/#351/#352 text is preserved, and the approved #357 additions are byte-identical. The two scripts and predicate test have unchanged blob hashes versus approved 98f27e3; range-diff cf6725d..98f27e3 vs 7ecb50f..fd2e2a5 shows context-only differences.

Exact-head remote Bun 1.4.2 checks: typecheck clean, boundary OK (170 source files), 14 focused tests / 334 assertions / 0 failures (255 ms), docs OK (5 stale stamps), pack dry-run 180 files / 731.3 kB. Verify exited 0 in 15 seconds. No full-suite rerun for this context-only rebase. Hosted CI on the new head remains unverified and required before merge.

Earlier implementation and head-attributed evidence:

Rebased onto main cf6725d at 98f27e3. The only resolutions are CHANGELOG and contracts context: all main #351/#352/#344 text and the exact approved #357 paragraph remain.

Both script blobs and the focused test blob are identical to approved ece9ea4. Range-diff a47710b..ece9ea4 vs cf6725d..98f27e3 changes only CHANGELOG/contracts context; there is no advisory, limit, predicate or product change.

Rebase checks on native Bun 1.4.4: 14 focused tests / 334 assertions / 0 failures (203 ms), boundary OK (168 source files), pack dry-run 178 files / 724.0 kB, diff-check and working tree clean. Remote final-head typecheck/full rerun was deferred because three other verify jobs occupied the project cap; no fourth job or heavy fallback was started. Earlier checks below remain attributed to their original heads. Hosted CI on this new head is unverified and remains required before merge.

Earlier implementation and validation evidence:

Refs #357.

Hosted Linux StopBurst runs can reach the existing 64-follow-up durability ceiling while spool backpressure leaves an unfinished transcript prefix. Apply the person-approved advisory only when that logical run contains explicit spool_full stops, a real positive cursor below the final expected-summary end, positive scanned-summary counts with every scanned key still durable, valid telemetry throughout, and zero R1 survivors. Missing or invalid evidence keeps the failure strict.

Warnings use the existing reportBudgetAdvisory log, annotation and mandatory step-summary receipt and contribute to incompleteRuns. Stop and SessionEnd durability, every progress/no-loss/RSS/R1/wall check, macOS behavior, limits, baselines, factors, MAX_FOLLOW_UPS and the workflow are unchanged. The contract and CHANGELOG record the scoped ceiling and retirement after a product fix gives 10 consecutive strict-green hosted Ubuntu runs, or a later person ruling. The issue stays open for AC3's hosted-run record.

Validation on remote Linux with Bun 1.4.2:

• Final head ece9ea4: typecheck clean, boundary OK (164 source files), 25 focused tests passed across 3 files (535 assertions, 521 ms), pack dry run passed (174 files, 719.4 kB), readme-facts OK and docs OK (4 existing stale documents). Verification exited 0 in 15 seconds.
• Full suite at code head 3a15521: 3527 passed, 25 skipped, 0 failed, 3552 tests across 314 files, 49244 assertions, 1316.34 seconds. Verification exited 0 in 1332 seconds. The final head changes only CHANGELOG.md and docs/contracts.md from that code head; code and tests are identical.
• Focused predicates cover each missing condition, invalid counts, Stop/SessionEnd rejection, exact hosted markers, unchanged macOS behavior, warning receipts, mandatory summaries and preservation of an existing failure.

Hosted CI and the post-change 10-run AC3 record remain unverified.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-10T11:42:02.055500Z	ece9ea4	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.