test: diagnose intermittent JSONL inventory catalog preparation failure
@JAORMX is already working on this.
Since Sep 8, 2026.
- Dominant language
- Go
- Stars
- 152
- Forks
- 16
- Avg merge
- 14h 48m
- Merged PRs (30d)
- 536
Description
Problem
TestUsableAutoSkillsStockBuildPolicyMatrix/explicit_evaluated failed on an otherwise unrelated CI-only perf-workflow PR with:
jsonlstore: inventory changed repeatedly while preparing catalog
Observed in PR #1222, CI run 34153750204. The PR changes only .github/workflows/perf.yml, so it cannot directly change the learning or JSONL-store execution path.
Relevant path
Store.readyInventoryCatalog makes up to three attempts to observe a stable inventory fingerprint around discoveryMetaListLocked, then returns the error above if the inventory keeps changing:
internal/adapter/store/jsonlstore/metalist.go:281-310internal/app/usable_auto_skills_matrix_e2e_test.go(TestUsableAutoSkillsStockBuildPolicyMatrix)
The test's diagnostics show the stale-session sweeper's list operation failed with this error, after which the expected learned skill was absent.
Related work
#1088 tracks a different intermittent failure in the same matrix test (learned skill operation failed) and suspected skill-store publication contention. This report records the distinct, actionable JSONL inventory-catalog failure path.
Acceptance criteria
- Reproduce or otherwise identify the concurrent inventory mutation that exhausts the three-attempt catalog preparation loop during this test.
- Determine whether the retry policy is insufficient or a test/production synchronization defect exists.
- Preserve fail-closed cursor and inventory consistency semantics.
- Add a deterministic regression test once the source is identified.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.