[#644 writing] Draft manuscript notes: P4 matching-odd signal / scalar-closure failures, P5 exact continuation / bounded-state no-go

[LightChainr]

First-pass writing drafts for issue #644 (ref, does not close). No new computation, no STATUS edits, nothing promoted.

What is here

• docs/manuscripts/p4-matching-odd-scalar-closure/README.md — Paper 1: the matching-odd finite-size signal and the failure of scalar closure. Signal existence (two independent blocks, block-diagonal contract), frozen H4 transfer vs aliases, frozen slope correction at N290; the three scalar-closure falsifications (multiplier 9.3520/2, width 24.5004/10, pure P4[S'] ~ N^-5/4 52.71634/2) as the contribution. Methodological spine: exact channel map DeltaS_cross = -DeltaS_either, finite Russo identity, frozen-prediction discipline. Explicit §4 correlated-block accounting (GOVERNANCE §2E) and the N=580 ladder replay: the covariance is now measured, not reconstructed (#575) #577 aspect-ladder limitation as a numbered limitation.
• docs/manuscripts/p5-exact-continuation-bounded-state/README.md — Paper 2: exact continuation, branching, bounded-state no-go. Part I cut-network theorems (r_d(k)=min(d+1,k+1), no bounded radius, ordinary-vs-nonnegative rank), Part II docs: #594 Q3 bounded-task-rank vs threshold no-go #615 no-go (p_c=1/2 vs 1/3, identical responses), Part III reflection quotient (block-for-block equality, pointwise Duhamel selection rule, exact factorization with inert odd sector). The ticket's settled boundaries are in the prose: Callan–Smiley/Burnside/Ding counts cited not claimed, w=4..8 lumping-orbit agreement called an instance (D'Angeli–Donno Prop. 13), P398 is a calibration process, not percolation stated in the abstract and its own section, not a footnote.

Deliverables mapping (issue #644)

1. two standalone drafts, neither claiming the other's ground — separate directories, disjoint evidence types, no cross-paper claims.
2. every number traced to a committed artifact path — each §-level number carries path → JSON field; the full-precision values in the evidence index at the end of each draft were re-read from the committed JSONs on main at drafting time (verification only, nothing recomputed).
3. explicit limitations with correlated-block accounting — P4 §4/§8, P5 §8.
4. list of what had to be left out — P4 §10, P5 §9.

Findings reported instead of produced (per the claim boundary)

• Note-vs-committed path drift: the docs: #594 Q3 bounded-task-rank vs threshold no-go #615 note's Files section names scripts/no_go_theorem.py / results/no-go/latest.json and the cutnetwork notes name scripts/rank_lower_bound.py / results/rank-vs-depth/latest.json; the committed locations are scripts/theory/… and results/cutnetwork-…. Drafts cite the committed paths.
• No references.bib entries exist for the Paper 2 cluster (Callan–Smiley, Ding, D'Angeli–Donno, Hänggi, Kemeny–Snell, Buchholz, Wonham, …) nor for Paper 1's FSS/crossing background. Bib patches deliberately deferred to a venue pass — detail is fully committed in the [Grok / deep retrieval] Symmetry quotients, selection rules and Fieller design: five things #598 may be rediscovering #601 packets and transcribing it here under a no-new-work ticket would skip the source-verification convention.
• P5 widths 9–10 (2494/8524) have no committed re-derivation — stated as a gap, not run.
• No generated tables.md/evidence-assembly script yet (p2/p3 pattern not established for these two) — stated as open item 1 in each draft's omissions list.

STATUS.md untouched; #644 stays open; no merge.

[LightChainr]

Verification comment, not a merge request. Ticket stays open.

Writing ticket as assigned: two standalone drafts, no new compute, no STATUS, no venue claim. P4 abstract carries the issue-body chi2 numbers and the three scalar-closure failures; P5 states in the body that P398 is not percolation and that Callan–Smiley/Ding counts are cited not claimed. Path-drift vs old notes is recorded rather than silently "fixed" by inventing files.

I did not re-read every JSON path cited in the evidence indexes. This is a scope check that the two papers exist, stay disjoint, and honor the claim boundary. Do not close #644.

[LightChainr]

Grok verify (independent of the CLI self-check). Not a merge. Ticket stays open.

Tripwire vs #644:

• Two standalone first-pass drafts, writing only; no docs/STATUS.md edit, no new compute, no venue claim.
• P4 headline numbers match committed artifacts on main: issue212 joint_scores χ² 31.18573555150965/4 and 3.4622795373044295/4; P57 H4/H12/H8 0.41630376 / 35.19307888 / 16.01202241; P50 joint ΔM χ² 9.35200368/2 (REPORT 9.35200/2); thermal-jet width 24.50042493/10; P48 pure N^{-5/4} 52.71634/2. Channel map 0.5700315436/2 is in the draft as assigned.
• P5 claim boundary holds in the body: P398 is not percolation; Callan–Smiley/Ding cited not claimed; D'Angeli–Donno Prop. 13 keeps w=4..8 as an instance. No-go witness: family1_pc=0.5, family2_pc=1/3, responses_identical=true. Reflection quotient blocks 10/26/76/232/750 at w=4..8 with coarsest_equals_orbit_partition=true. Widths 9–10 left as un-run.
• Nit only: P4 NLPD last digit is −13.806418789729096 in the draft vs −13.806418789729094 in issue212-matching-odd-synthesis.json (2 ulp; not a claim change).
• Does not close [P1 writing] Two finished results are sitting unwritten while we wait for a unified theory that is not coming #644.