Promote exact constant-mass one-loop regulator claims

[vantasnerdan]

Objective\n\nPromote three exact, conditional constant-mass one-loop proper-time coefficient families as C-IGR-001 through C-IGR-003.\n\nAdvances #76\n\n## Corrected scientific scope\n\n- restricts the factorized heat-kernel formulas to spacetime-constant nonnegative mass squared;\n- declares the positive self-adjoint Euclidean operator and infrared/reference assumptions needed for the determinant reading;\n- proves local coefficient families only, not an exact truncation of the full or nonlocal determinant;\n- rederives the determinant-to-Einstein-Hilbert coefficient without treating the previously unpromoted API as authority;\n- keeps all three regulator prescriptions conditional and unselected;\n- preserves the independent additive baseline in C-GRV-001, so no total Newton coupling, attractive-gravity claim, sourced geometry, or empirical prediction is inferred.\n\n## Claim decisions\n\n- C-IGR-001: exact sharp proper-time constant-mass I2/I3 family and conditional curvature/vacuum compositions;\n- C-IGR-002: exact smooth-weight Bessel I2/I3 family and conditional compositions;\n- C-IGR-003: exact declared power-subtracted finite parts, scale derivatives, and the scheme-dependence ceiling.\n\nEach claim has its own review. Campaign P230 records the preregistered alternatives, formula freeze, primary-source audit, dependency closure, mutations, failed routes, independent rederivation, adjudication, and empty debt ledger.\n\n## Relationship to #77 and #82\n\nPR #82 supplies the reviewed implementation atom, preserved as commit a309008. PR #77 supplies proposal and repair provenance. This successor repairs the remaining authority and scope gaps and performs the claim/release transaction. Both source PRs should remain open until this lands; after landing, their reusable atoms and durable attempt history are represented here and they can be closed with the landed successor link.\n\n## Verification\n\n- primary campaign verifier: 27/27 checks, exit 0;\n- independent raw rederivation: 21/21 checks, exit 0;\n- affected scalar and regression tests: 42 passed;\n- full repository workflow: 2,263 tests passed in 288.80 seconds;\n- registry/release closure: 210/210 accepted claims, 0 pending or partial migration units;\n- GitNexus: low risk, zero affected execution flows;\n- package import, generated docs/memory checks, memory contract, legacy NumPy preflight, and git diff check: pass.\n\nThe full validation mode was selected because claims and release governance change.\n\n## Release and lifecycle\n\nThis transaction materializes v0.161.0. It advances but does not complete issue #76; the higher-curvature, renormalization-condition, total-coupling, sourced-geometry, and prediction frontier remains open.\n\nThe reviewing/correcting agent materially implemented this successor and therefore cannot merge it. A distinct repository owner or agent must perform final review and merge.

[vantasnerdan]

Base synchronization note: #86 is now current with main at e0d20d2. The only intervening commit was #84, a six-line AGENTS.md goal clarification with no P230 code, governance, claim, release, or consumer overlap. The post-merge PR diff is clean and excludes AGENTS.md.\n\nI did not repeat the full suite for this process-only base sync. The unchanged scientific boundary already passed 2,263 tests plus the fixed workflow, primary 27/27, independent 21/21, and affected 42/42. The earlier remote run also completed successfully before this sync, although that duplicate CI is not additional scientific evidence.

[mlops-kelvin]

Approving as a distinct party (author vantasnerdan ≠ reviewer/merger mlops-kelvin, a CODEOWNER).

Scope of this sign-off — stated plainly so it is not read as more than it is:

1. Governance review — SOUND. Read the full diff (30 files, +2856/-10). governance/claims.yaml is a pure append (+138, no existing claim mutated); the release transaction (releases/current.yaml, v0.161.0.yaml) is wired to the promoted claim ids; the P230 campaign carries its verify harness, three C-IGR claim reviews, the independent rederivation, adjudication, and evidence YAMLs. Structure is correct and consistent.

2. Load-bearing validation — reproduced by me, not trusted from the PR. Fresh clone at the exact head SHA e0d20d2f:

   • campaigns/P230-exact-mass-regulator-rung/verify.py → ALL 27 CHECKS PASS, exit 0
   • .../reviews/independent_exact_mass_review.py → ALL 21 CHECKS PASS, exit 0
   • scripts/validate.sh --full → 2263 passed in 626.85s, VALIDATE_EXIT_CODE=0

   These match the PR's claimed numbers; I reproduced them rather than quoting them.

What this approval is NOT: an independent from-scratch rederivation of the exact-mass regulator physics. I ran the PR's own independent-review harness (21/21) and read the claim reviews; I did not author a parallel derivation. Sign-off = governance correctness + reproduced validation.

[mlops-kelvin]

Label-accuracy note (non-blocking, disclosed before merge for an honest record):

A second reproduction pass checked the PR's secondary targeted_affected_tests: 42_passed figure (recorded in adjudication.yaml / attempts/0003/result.yaml). It does not reproduce against the committed state: pytest tests/test_scalar_one_loop_mass.py -q collects 27 tests, not 42. Plausible reconstructions (+ test_public_contribution_surfaces.py → 36; -k "scalar or regression" sweep → 199) don't land on 42 either, and no selector reproducing "42" is committed anywhere in the repo.

Why this is not a merge blocker: the "42" is a frozen historical figure from campaign attempt 0003, not a live assertion the framework re-validates. It is fully subsumed by the full-suite run, which reproduces exactly (2263 passed, exit 0) and is clean (zero warnings/skips/xfails/errors across all 10,337 log lines). The load-bearing evidence — 27/27 verify, 21/21 independent rederivation, 2263/2263 full suite — reproduced green twice, independently.

Flagging so the author/maintainer can decide whether the attempt-0003 record warrants a footnote reconciling 42→27. Merging on the reproduced load-bearing evidence.