[KevinBigham]

The main release gate on 463903f6c3139a674f2f982dd39fc54b123e45a3 timed out waiting for Ivy Patch in staff-facility-medical. The failure page showed the original regular-season Week 14 save with no medical staff, rather than the staged offseason fixture.

The successful PR #108 integration and failed main commit share the exact tree 787f2cba8dfca5ffda9818445ab0d8f28254c4a2. The scenario's IndexedDB helper resolved on put request success, before transaction commit, and immediately hard-reloaded. A controlled regression using the actual browser expression reproduces premature completion and false success after a subsequent abort. This supports an intermittent fixture commit race; the historical log did not record an abort, so that particular event remains inferred.

Correction

Wait for the medical fixture transaction's oncomplete, reject request/transaction failures, and close the connection on write failure. Add four behavioral regressions for commit ordering, abort after request success, transaction error, and request error. All existing candidate, facility level/effect/budget, hire/pool-return, hard-reload and browser-error assertions remain intact.

Verification

• Before correction: new tests exposed premature completion, missed abort/error and missing failure cleanup (3 failed, 1 timed out).
• After correction: all 83/83 configured Node script/release-contract tests passed, including all four regressions. Syntax and diff checks passed.
• Independent read-only review approved the focused change and independently passed 74/74 relevant contracts. It checked import awaiting, fresh Continue reads, and passive route behavior; additional fixture isolation is unsupported by current evidence.
• Source publication tree matches the tested local tree: 28fd0ed3e01e1376fcd6de2affd5f14f26cf3ebb.
• Hosted CI passed all three required jobs in run 37806896603, attempt 1, on head 8262621f19819b07adcadeef7bf3fcffbf3f8dd9. The integration commit 3b0951c5a3676d7d86f9165743a03ec210548c8e has the same tested source tree.
• The workspace test job passed 4,854 tests across 534 files. The complete release gate passed 37/37 steps in 2,634.4 seconds, including all 15/15 G3 workflows. The corrected medical scenario completed in 6.7 seconds, upgraded the facility, hired Ivy Patch, verified persisted effects/budget/pool return through the unchanged hard reloads, and reported no browser errors.
• Main must pass the same full gate before the existing Pages workflow can deploy its exact SHA-named artifact. The local environment has no Chrome; native browser evidence above comes from the unchanged hosted release job.

No application, save/schema, gameplay, dependency, CI job, release-plan, or deployment changes. Reverting these two files restores the old test helper.

Failure evidence: https://github.com/KevinBigham/MFD/actions/runs/37723894338/job/113143484567
Passing identical-tree evidence: https://github.com/KevinBigham/MFD/actions/runs/37716671343/job/113123157060

Main validation

Merged as d617b691342b4bd552a43edc6a929150180f635b; its tree matches the successful PR integration and locally tested correction exactly. Main CI run 37817114932 is pending. Pages remains gated on successful completion of that main run.

[chatgpt-codex-connector]

Codex Review Summary

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

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-08T16:13:02.216835Z	8262621	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.