chore: iss-24 merge-window evidence and the run's six deferred captures

[REPPL]

Post-run ledger reconciliation, records only — two commits.

iss-24 evidence (the changelog-conflict answer)

Appends the merge-window evidence to iss-24 (changelog fragments): three server-side CHANGELOG conflicts in one day while the run's eleven PRs merged, each resolved by a manual merge-main-into-branch because the merge=union stopgap only works locally — the hosting platform's server-side merge ignores custom drivers. Also records the design constraint discovered since capture: the release gate now derives the release tier from changelog records (releaseImpact, landed with iss-122), so fragment assembly must preserve the impact-derivation seam. Recommendation recorded in the issue: grill-then-implement, queued at the top of the next run.

The six deferred captures (iss-132 … iss-137)

Held back deliberately until the id-allocator class fix (#137) reached main — minting on the run's branches would have reproduced the exact iss-115/120 collision the run fixed. Now minted sequentially by the refs-union allocator:

• iss-132 — marker scan: three irreducible quoted-delimiter doc false positives (iss-111 residue).
• iss-133 — probe truncation message overstates scope (conservative direction, but imprecise).
• iss-134 — readDirBounded determinism doc claim only holds at or under the bound.
• iss-135 — WalkFiles non-dot start divergence (unreachable today; flagged for future callers).
• iss-136 — itd-88 audit: the promised Pass-B exemption marker in _provenance.json has no implementing code.
• iss-137 — itd-88 audit: probe/coverage ship as CLI verbs only; the plugin surface doc stops at plan/pack.

Verification

record-lint exit 0 and make preflight exit 0 after all edits.