[amaybaum]

Source-correction unit SC-22 of the #942 audit (gate #946), carrying the owner's three decisions. Base: main 0d917665. Head: 621fbf2c.

Review history:

Commit	Content	Review
eaeed819	First draft	ACCEPT WITH CHANGES (R1–R4; ch18 :86)
621fbf2c	Applies R1–R4, ch18 :86 and the optional findings	Confirmation: ACCEPT

Owner decisions:

• (i) 𝒩 is the peak revival. On a window W, 𝒩 is the largest [d_t − d_s]⁺ over pairs of preparations and times s < t in W, where d is the total-variation distance of the visible population laws (equivalently, the trace distance of the diagonal states). It lies in [0, 1].
  • The Breuer–Laine–Piilo sum is named separately, as cumulative backflow. It is unbounded on a finite bijection, because Γ_{t+L} = Γ_t [Main §2.3].
  • 𝒩 > 0 certifies P-indivisibility; the converse fails.
  • Memory does not fix 𝒩: the XOR control has memory on a window where its family is P-divisible.
  • The anchor is ch18 §18.3, mirrored in the glossary.
• (ii) Relabel. "Partially-quantum" becomes "restricted quantum-statistical regime" at 57 sites, each mirrored in FULL. The anchor states that the restriction is to the transition-statistics layer, not a complete operational quantum theory.
• (iii) Table. The ch18 :82–:88 table is removed. Three conclusions survive in prose:
  • a fast bath with uniform erosion drives the readback gap to zero (Main §2.3, τB ≪ τS);
  • a static bath gives no memory and so no revival;
  • the candidate systems, labelled as the proposal's conjecture.

Review corrections:

• ch15. F = 1 is "saturated revival, 𝒩 = 𝒩_max", and the regime is the sub-saturated case. A non-monotone G² is the proposal's prediction and does not certify a revival. A P-divisible chain with 𝒩 = 0 can have an oscillating G². All distinguishability uses total variation.
• ch09 :285. It now says "marginal capacity (N_H/N_V ∼ 1)".
• ch18 :86. It now reads "a memoryless (divisible) density-matrix description requires …". This is consistent with universal representability, and is the wording at ch13 :144.
• "C3 marginality" becomes "marginal hidden capacity", since r = N_H/N_V is not Main's C3 quantity.
• E4 is narrowed to N_H/N_V ∼ 1.

Evidence (§A.42). The record's numbers rest on six probe files, landed byte-identical to the audit branch at 42bbd33c under tracks/SC22-N-regime/: regime_exact, blp_unbounded and enum_window, each with its .py and .out. Each replays its output under python3 -I -B.

Excluded and recorded on #946:

• ch01 :225 and appendix-c :62 still use the old label; their owning units will relabel them.
• The "specific algebraic predictions" at ch13 :138–:170.
• The 𝒩 symbol collision with SM.
• The architecture_check label patterns.

Introduction dependency: none.

Checks (local, at 621fbf2c):

• Release gate: every step passes except lean-axioms, which runs in CI.
• edge_rigidity_probe: all checks pass.
• Mirror exact; the FULL stamp matches.
• Book: 558 pages.
• Added-lines and secrets scans: 0 hits.

Merge: in dependency order after #918, and only at an exact SHA the owner approves.

Links: #942, #946.

---

Generated by Claude Code

[amaybaum]

Closed unmerged. All 156 hunks are on main: 144 verbatim, 2 deletions and 10 kept with later refinements. They landed through #1019 D-5 (daa62705, f834ef37, 869f4ddc, 826f34c6); see also D-8 and #1035.

The three exclusions this PR recorded are carried to open issues (owner decision OD-8):

• (1) ch13:142/:146 "specific algebraic predictions": Correction planning: scientific overclaims and status corrections (from the 2026-11-01 disposition) #1028 (comment 6597310497);
• (2) the 𝒩 symbol collision and (3) the architecture_check.py label patterns: Correction planning: cross-references, labels, ledger/census and evidence records (from the 2026-11-01 disposition) #1029 (comment 6597316771).

The SC22 probes are not needed by any claim on main. They stay on this branch (regime_exact.py f65a9286, blp_unbounded.py 418a96ed, enum_window.py 53581d50). Branch corpus/sc22-n-regime is kept.

---

Generated by Claude Code