Revenue read-only shared release inventory
Assignment ADR-0055/revenue/003, coordinator 01a0e095-f279-7fd0-a8ef-92ce8f9547fc/local, participant 01a0ff07-dae9-7ed3-9b27-0fd2c4dcc6fe/local. Bounded read-only inventory is complete. Root is sole canonical publisher/indexer. No production writes, migrations, role changes, flags, build/deployment or provider HTTP occurred. Frozen runtime 127d59b59ab745fb777ee20f3e04237c287cbd21, authentic original and old listeners remain unchanged. This memo neither changes other ADR ownership nor completes shared release activation.
Current canonical Revenue production is Ready/promoted deployment dpl_DjpzM72TiEFko4BUBXFmjDG19rTJ, explicit release source 5417af4cf384ecc91f8a1da7a671a4094993ee05. Checkout metadata ec4665b25a01e575c15983f74569e326cd12b5ff adds only documentation. The released identity correction and test are already identical blobs in the frozen candidate. Actual serving build is npm run build, Node 24.x; candidate's implicit migration-first build and alternate-schema data-loss branch require a deliberately separate migration/build plan in a new release candidate.
Fresh explicit repeatable-read/read-only catalog inspection of the prior ADR-0049 release-bound Supabase project pfuctqiobbjglpukteei, database postgres, canonical schema revenue, finds 47 tables, 65 enabled noninternal triggers, six functions and 51 migration journal rows: 49 completed, two historical rolled-back attempts. No unfinished attempts, unknown journal names or source checksum mismatches. Historical entries and trigger definitions are unchanged from the retained post-correction catalog. Of the 14 pinned inputs, the two September closure/handoff migrations are already applied, leaving 12 pending. The catalog and exact byte-representation comparisons are pinned in the Revenue packet.
The catalog inspector is elevated postgres, with BYPASSRLS, role creation and relation ownership. It is not genuine serving-role evidence. Current Vercel credentials withhold sensitive database URLs, so current application target equivalence and actual role restrictions remain unproved. Safe current project flags are fanout true, handoff true, both onboarding closure/provider flags false; immutable full runtime environment is not attested.
The writer inventory covers 34 guarded tables, 137 candidate writer/sender files, 75 API entries, 13 families and all five cron paths. Closure epoch 1 is composed globally in the candidate client, but ordinary standalone writers do not all enter the scheduling epoch marker. Before scheduling barrier enablement, integrate the relevant paths or explicitly pause them, then prove admission and drain executable old provider/retry/cron/operator paths. Distinguish executable PG passthrough from deterministic retired 501 routes. The disabled local read-only proof role cannot stand in for the production application writer role.
The packet provides the exact 12-migration canonical ordering, schema-only rehearsal/preflight and post-apply checks, effective role/ACL requirements, compatible binary rollback retaining permanent facts and barriers, and a concrete future bounded diagnostic method for the actual serving connection plus deployed raw-header/auth/owner proof. No production apply or diagnostic execution is authorized by this report. Public immutable dispatcher 2.0.3 packaging and Square owner login remain external prerequisites. Separate active Square changes retain their owner.
Evidence is under C:/Users/jallr/.codex/worktrees/revenue-adr0055-observation/revenue/docs/adr0055/release-readiness/, with evidence-index.json providing exact artifact byte pins. The source integration inventory also retains primary dirty-file hashes without changing those files. No new implementation tests were needed for documentation/catalog-only work; the frozen implementation's previous test evidence remains the source validation basis. Root should publish this owner-authored memo with the exact packet index and retain all named release gates.