[LIghtJUNction]

Refs #81, #82, #83, #84, #85. This integrates the reviewed partial fixes, not completion/closure of the umbrella issues.

Included commits (preserved as parents)

• fix(registry): prevent schema aliases from changing frozen contracts #95: frozen Registry schema ownership; no aliases may change validators or published contracts.
• fix(tui): add safe read views, narrow-terminal layout and explicit recovery #96: safe read-only TUI views, terminal cell wrapping, explicit reconnect/current authorization and secret-safe status.
• fix(backup): retain money policy during isolated recovery #97: preserve money policy during isolated restore; no production DSN/mail/cache activation.
• refactor(effects): share retry transitions without weakening lease fencing #98: behavior-equivalent fenced mail/Webhook retry transitions.
• ci: fix nested fixture collection and require complete sharded evidence #104: remove conflicting fixture imports and verify four complete disjoint pytest partitions at the existing contracts gate.

The application/test blobs are byte-identical to those individually reviewed PRs. The merge tree starts from current main 2dd6a0d0260f8c230a1d5b2f4ff75241113835b2, preserving parallel #101 and earlier journal work. #93 and #94 are already on main and also remain present.

Why an integration PR

The previous sequential 20-minute CI jobs for #95/#97/#98 were cancelled before complete evidence; a failed #96 run must not be treated as green. The new CI gate verifies the actual combined code and retains all test nodes, real PostgreSQL/Valkey/Python 3.15, tokenizer conformance and installable builds. Nothing is merged based solely on prior targeted tests.

Local evidence so far

The six original changes passed 100 combined targeted tests on Python 3.13.5 / PostgreSQL 16.15. The additional collection/CI evidence suite passed 18 tests. A full latest-main integration run is separately recorded; results will be added only when complete. Local Redis is not evidence of real Valkey acceptance.

Merge / boundaries

Use a regular merge to retain each reviewed parent so the constituent PRs are included in history. Do not close #81–#85: notification preferences/public delivery/collaboration primitives, full CLI/TUI writes and real-client isolation matrices, full feature/config/doctor coverage, legacy migration/production recovery/capacity and remaining HTTP/journal/ledger refactors remain unaccepted. No production deployment, production restore, widened authority, schema migration or automatic external notification activation.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
🔒 Security Review	✅ Completed 2026-09-27T20:57:36.646031Z	94c58e7	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[LIghtJUNction]

Latest local integration evidence (2026-09-27 UTC): all application patches on pinned main 2dd6a0d plus the fixture collection repair completed pytest tests: 1029 passed, 2 skipped, zero failures, 716.54s. The skips are the real git-lfs client and age-keygen interoperability tests because those binaries are absent locally. The additional CI partition/collection suite separately passed 18 tests; these CI-specific tests were added after the long run collected its items, so do not add the counts without accounting for overlap.

Conformance: 7 passed, 1 failed. The six real HTTP/path-GET/GraphQL/MCP-HTTP/CLI/MCP-stdio transfer cases and actual cl100k_base tokenizer check passed; the sole failure is the unchanged release assertion requiring Python >=3.15 while this local interpreter is 3.13.5. The assertion has not been bypassed or weakened. Local build initially cannot import setuptools, an unavailable local build dependency; this is not a successful build claim. The PR's configured Python 3.15 CI remains the required gate.

Correction to the earlier preliminary CI description: fresh GitHub run/job metadata reports all four old #95–#98 jobs as cancelled, not a verified successful run. A previously retrieved connection-exhaustion log is not a confirmed root cause for this head; no connection-exhaustion failure occurred in the complete local core run. Do not label that as a fixed defect. No production deployment.

[LIghtJUNction]

Local build follow-up: using the already-installed setuptools 82.0.1 from the isolated environment, python -m build --no-isolation now succeeds for both sdist and wheel. Wheel metadata was inspected: Requires-Python: >=3.15 is unchanged and only the intended msg / msgd console entry points are present. This remains a local packaging check, not Python 3.15 runtime release evidence.

Final current-tree collection is 1048 core test nodes (the earlier 1031-node complete run plus 17 newly added CI-evidence cases; the separate 18-test check also repeats the one collection regression). JUnit confirms the earlier complete run had zero failures/errors and exactly two documented tool-availability skips.

[LIghtJUNction]

Reviewed the integrated changes against current main and the built source artifact. Registry keeps owned schema snapshots; recovery retains money configuration without copying outbound services; retry refactor preserves state/attempt fencing and channel codes; TUI changes remain read-only with explicit retry and secret-safe rendering. The CI collection fix removes redundant imports rather than replacing the real PostgreSQL fixture. All four exact-head shards and the contracts evidence gate succeeded in run 36349643025; shard 0 also passed all 8 conformance checks and built wheel/sdist. The manifest contains 1,048 collected tests. This validates this partial patch set only, not the full #81–85 acceptance scopes or any production deployment.