[freeqaz]

Problem

The flake-sweep-1020 window (10-19 18:00Z to 10-20 04:18Z) found five shared files with no row in file-durations.json: console-namespace-create, console-shell, console-page-header, credential-source-api and agent-create-validate-api. A file without a row starts first in its lane, ahead of longer measured files. console-focus-ring's row was 11.1 s, but #2977 grew the file and it now measures about 15 s.

Change

scripts/ci/test-suites/file-durations.json gets five new rows and one refreshed row. Each row uses the table's documented method: the slowdown-normalised median over the green main runs whose head contains the file's last change. Each run lane's factor is the median, over its files other than these six, of their time divided by their current row. The source text records this.

File	Row	Runs
console-namespace-create	15.6 s (new)	3 (main run 81c57a7 and #3002's two branch runs, after #3002 grew it from 7 to 10 tests)
console-shell	12.8 s (new)	2 (last changed by #2996)
console-page-header	8.5 s (new)	15
console-focus-ring	11.1 -> 15.5 s	15
credential-source-api	4.2 s (new)	21
agent-create-validate-api	3.2 s (new)	21

Data: the sweep's downloaded ci-results-<lane> artifacts from main runs in the window, plus #3000. console-namespace-create is the exception: #3002 grew it after the window, so it uses the runs listed in the table. The branch is merged with main.

Verification

• node scripts/ci/run-tests.mjs audit passes. It validates the duration table along with the suite manifests.
• The ci-runner duration and start-order cases pass.
• Lane balance (longest-first greedy over fileConcurrency slots, using the new times): Console Browser 88.6 s, Console Browser 2 90.3 s, Checks and Conformance 1 54.4 s, Checks and Conformance 2 56.5 s, unchanged from the old start order within 0.1 s. The new rows move the files out of the "unmeasured, start first" group: Console Browser goes from 3 unmeasured files to 1 and Console Browser 2 from 1 to 0. The one left, console-create-agent-member, landed in fix(console): let members with agent:create use the create form #2998 after the window.
• prettier --check passes.

Risks

• Start order only. No test changes. The console-shell row comes from only 2 runs and the console-namespace-create row from only 3, so they are the least certain.

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.

ClawSweeper review complete

ClawSweeper finished reviewing this revision. The review result is being finalized.

View the workflow run.

[clawsweeper]

Codex review: needs maintainer review before merge.

What this changes

This PR adds five measured test-file durations and updates the console focus-ring estimate.

Example: Scheduling console-shell.test.mjs in Console Browser

• Before: The missing estimate places it ahead of every measured shared file.
• After: Its 12,800 ms estimate places it behind longer measured files.

Review scores

Measure	Result	What it means
Overall readiness	🦐 gold shrimp (3/6)	The patch is focused and supported by existing integration coverage and successful exact-head checks; its measurement calculations and claimed lane balance were not independently demonstrated.
Proof confidence	🦐 gold shrimp (3/6)	This collaborator-authored CI metadata refresh is outside the ordinary contributor proof gate; the inspected console screenshots concern linked UI work and do not prove scheduling behavior.
Patch quality	🐚 platinum hermit (4/6)	No actionable review findings were identified.

Product

Kind: Maintenance · Worth it: Yes · Fix scope: Complete
User problem: CI scheduling estimates become incomplete or stale as test files are added and expanded.
Reason: This maintainer-authored refresh follows the existing scheduling design with a small metadata-only change.

Merge readiness

✅ Ready for maintainer review

Keep this PR open for maintainer triage: it is a focused refresh of CI scheduling metadata, with no blocking correctness or security finding.

Priority: P3
Reviewed head: cf659b7b1baacd28bd23ec59d23cfad798aeaed6

Before merge

None.

Findings

None.

Agent review details

How this fits together

The CI runner reads duration estimates and suite manifests, orders eligible shared files, and executes them within the lane's concurrency limit.

flowchart LR
  A[Suite manifests] --> C[CI runner]
  B[Duration estimates] --> C
  C --> D[Shared file ordering]
  D --> E[Concurrent test execution]
  E --> F[CI results]

Loading

Technical review

Best possible solution:

Keep measured scheduling estimates current within the existing table and runner.

Do we have a high-confidence way to reproduce the issue?

The current-main table and runner establish the missing-estimate ordering directly; no target code was executed during this read-only review.

Is this the best way to solve the issue?

Refreshing the existing table is appropriately scoped, although the credential-source-api estimate is currently inert because that file runs serially.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning medium; reviewed against 8fe865c1649a.

Provenance checked

• Measured longest-first scheduling keeps the original intent (ci(tests): start shared lane files longest first from measured times #2655: Avoid leaving long shared files waiting behind short files and refresh estimates as tests change.)
• Console focus-ring duration keeps the original intent (ed489de: The earlier console-test consolidation adjusted estimates for moved tests; the later file growth warrants remeasurement.)

Testing

Proof path: in-process harness.

Security

None.

Evidence

What I checked:

• Verified introduced change: The pinned introduction diff changes only the duration table: five added estimates, one refreshed estimate, and measurement provenance. (scripts/ci/test-suites/file-durations.json:28, cf659b7b1baa)
• Current-main scheduling contract: Current main still ranks unmeasured shared files first and measured files by descending duration; its table lacks the five additions and retains the 11,100 ms focus-ring estimate. (scripts/ci/run-tests.mjs:600, 8fe865c1649a)
• Existing integration coverage: Existing runner tests exercise measured and unmeasured start order through the CLI and reject invalid duration tables; this PR adds no tests or scaffolding. (tests/integration/ci-runner.test.mjs:3706, cf659b7b1baa)
• Original design intent: ci(tests): start shared lane files longest first from measured times #2655 introduced measured longest-first scheduling and explicitly anticipated refreshing stale rows; this PR follows that design. (scripts/ci/test-suites/file-durations.json, e834adc32de6)
• Exact-head checks: Suite Audit, both Console Browser lanes, both Checks and Conformance lanes, and CI Required succeeded for the reviewed head; no diff-caused failing check was identified. (cf659b7b1baa)
• Scope caveat: credential-source-api is listed as a lane file but is absent from parallelFiles, so its added estimate currently has no scheduling effect. (scripts/ci/test-suites/checks-baseline-1.json:350, cf659b7b1baa)

Labels

Label changes:

• add P3: This is low-risk CI scheduling maintenance without changes to application behavior, test assertions, or test selection.
• add rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🦐 gold shrimp and patch quality is 🐚 platinum hermit.
• add status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR.

Label justifications:

• P3: This is low-risk CI scheduling maintenance without changes to application behavior, test assertions, or test selection.
• rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🦐 gold shrimp and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR.

Rating scale

6/6 🦀 challenger crab · 5/6 🦞 diamond lobster · 4/6 🐚 platinum hermit · 3/6 🦐 gold shrimp · 2/6 🦪 silver shellfish · 1/6 🧂 unranked krab. Overall follows the weaker of proof and patch quality; ✨ marks media proof (a screenshot, video, or linked artifact) that directly shows the changed behavior.

Workflow

ClawSweeper edits this one comment on every review. Comment @clawsweeper re-review for a fresh review only. Repair and merge execution lanes are retired.

Reviewed October 20, 2026, 1:35 AM ET / 05:35 UTC.