[DavidWallstructurallaw]

Phase 6A is complete for development version 0.1.0.dev5 and report schema 1.2.

Explicit ordered comparison now covers adjacent and optional first-baseline pairs, observed distribution and earlier-tail disappearance, provenance/direct-closure changes, shared selected-target lineage, privacy-preserving JSON/Markdown and two packaged examples. Missing, unknown, zero and unavailable evidence retain their distinct meanings.

Tested candidate: 157109717079c9db9f02ac65ec1493709f37d8a4, tree 770aff9e12342d0ec0977d989bf3743acf535085.
Administrative handoff: b6389c6c50f2fc61d39580274bd24ed39e09ca45. Only the completion, plan and decision status documents changed; the other 320 tracked entries retain the candidate's exact Git identity.

All 13 required candidate jobs passed in https://github.com/DavidWallstructurallaw/recursive-integrity-toolkit/actions/runs/36327281529:

• Eight Ubuntu/Windows, Python 3.11/3.12, current/minimum core profiles: 3,622 tests each, with actual PyArrow absence.
• Real Parquet: 3,632 tests, including exactly ten optional cases; Hero: 302; security: 2,010.
• All 19 reference performance cases passed. The actual three-version 100,000-total-record longitudinal CLI took 965.342692 seconds, peaked at 7,340,011,520 bytes RSS and published 694,462,789 bytes of JSON/Markdown.
• Same-candidate delivery passed: installed wheel, safely extracted sdist execution, 41 modules/14 resources, exact 323-file source archive, identical repeated wheel bytes and identical sdist file payloads.

The first candidate 09fe89bb passed the matrix and longitudinal 100k case but failed two retained metadata tests because one oracle omitted the required empty schema 1.2 longitudinal family. Its failed tests and all measurements remain disclosed. The correction changed only that helper, retaining scientific assertions and requiring not_requested; the corrected candidate then passed the entire workflow. Local preflight failures and their focused rechecks are also retained in the completion record.

Full results, both measured attempts, comparison limitations and exact artifact hashes: PHASE_6A_COMPLETION.md.

This draft remains based on the accepted, unmerged Phase 5 branch. Phase 6B, main merge, tagging and package publication remain separate decisions. No stable release or empirical model-performance claim is included.