Release replay clarification

The first final replay stopped at C1424 because its link-report list is populated by iterating a Python set. A new process can serialize the same complete links in a different order. The 299-action preseal replay had matched the recorded order and passed all 774 checks.

The completed C1424 script, journal record and generated artifact remain unchanged. The original verifier also remains unchanged. verify_release.py adds one narrowly scoped presentation-order comparison for that artifact: it verifies the recorded artifact's exact hash and byte length, regenerates the report from the ZIP, and compares both JSON objects after sorting only local_links. All link values, multiplicities, other report fields and action checks must match. No numerical result, source identity, domain condition or other artifact comparison is relaxed.

Use python3 evidence/verify_release.py --through 1431 for the final portable replay. The verification report discloses this normalization. The correction is a release reproducibility fix, not a new numbered research action. No C1432 research was performed.