Tests: Shuffle test cases in CI

[ciampo]

Follow up to #80855.

What?

Shuffle tests inside each file as well as files in the existing Node/jsdom and Chromium CI jobs. Use a different seed for each new workflow run.

Why?

File shuffling alone cannot expose shared state between tests in the same file. Varying the seed exercises more orders, while retaining it makes failures reproducible.

How?

Enable both shuffle flags and pass GITHUB_RUN_NUMBER from CI. All shards and Chromium share the seed; reruns retain it. Each job logs its seed and reproduction command. Local runs use Vitest's current-time default unless an explicit --sequence.seed is supplied. Update the command policy and contributor guide to match.

Repair three setup dependencies exposed by seven fresh random seeds: initialize the clipboard before installing spies in Compose unmount tests, mark the seeded core-data record as resolved before editing it, and initialize the VIPS worker module within the cancellation test group. The cancellation tests also verify worker delegation when it returns false.

Testing Instructions

1. Run npm run test:unit:vitest:shuffled -- --sequence.seed=12345 --project=node --project=jsdom --shard=1/4. Repeat for shards 2/4, 3/4, and 4/4 on Node 24 and Node 26.
2. On Node 24, run npm run test:unit:vitest:shuffled -- --sequence.seed=12345 --project=browser.
3. Change the seed to exercise another order. To reproduce a CI failure, copy its logged command and use the same checkout and Node.js version.
4. Check that a new CI run logs a new seed and a rerun retains the original seed.

Verification

All 63 matrix cases passed on a5f05d10e2d60111aa7e9d3fc339ad912514c523, based on trunk 72e0cc4f894dde390838f64e588cc25b294c0fa3, with no failures or retries in the final sweep. Each seed ran all four Node/jsdom shards on Node 24.20.0 and Node 26.9.0, plus Chromium on Node 24.20.0. The seven seeds were drawn without replacement: 1244467973, 518844352, 317867813, 189184704, 372323832, 1484297706, 2139874380. Local runs used macOS arm64, two workers per Node shard, and separate Chromium runs.

Every Node runtime/seed passed 35,185 tests; every Chromium run passed 121 files and 1,680 tests. Existing expected-failure cases and skips were unchanged.

Each setup defect was reproduced before its repair. All four repaired files passed all seven seeds on both Node versions. The clipboard and VIPS failures also passed individually after failing alone before the fixes. Build, type checking, lint, test conventions, routing, and independent adversarial review passed.

Unit CI also passed on this commit. All nine shuffled jobs logged and used seed 179071; the other GitHub checks passed or were intentionally skipped.

All 58 infrastructure tests passed with explicit and automatic local seeds. A shell probe checked 27 actual workflow-command invocations across runtime/shard combinations, new run numbers, and a rerun, verifying seed forwarding and reproduction commands.

Prerequisite repairs and verification limits

Merged prerequisites cover mutable fixtures and registries in #83098, #83099, #83104, and #83107; UI setup in #83101 and #83103; RTC timer teardown in #83366; LinkControl readiness in #83489; and Browser iframe lookup in #83495. The timer, readiness, and iframe repairs address races, rather than proven dependence between test cases.

An earlier sweep on c66b6a20 hit a native Node 26/V8 memory abort at seed 189184704, shard 4/4. The exact command passed on isolated replay without source changes. Its cause remains unexplained.

Two earlier diagnostic Browser runs at seed 80856 reported eight other timeouts in Composite, Popover, Autocomplete, SearchableSelect, PaletteEdit, and FontSizePicker tests. Their cause remains unknown. The saved logs and current cleanup did not justify a repair; subsequent passing runs do not establish that those timeouts are fixed.

No assertions are weakened, and isolation, timeouts, retries, and skips are unchanged. No finite set of passing seeds proves order independence.

Testing Instructions for Keyboard

Not applicable. This PR changes test infrastructure and test setup only.

Use of AI Tools

Codex rebased the branch, implemented the seed changes and test repairs, performed independent adversarial reviews, and ran verification.

[github-actions]

🤖 PR meta 🤖

📦 Bundle size

Size Change: -16 B (0%)

Total Size: 8.22 MB

📦 View Changed

Filename	Size	Change
build/scripts/editor/index.min.js	613 kB	-16 B (0%)

_{4912a1d Run}

[github-actions]

The following accounts have interacted with this PR and/or linked issues. I will continue to update these lists as activity occurs. You can also manually ask me to refresh this list by adding the props-bot label.

If you're merging code through a pull request on GitHub, copy and paste the following into the bottom of the merge commit message.

Co-authored-by: ciampo <mciampini@git.wordpress.org>
Co-authored-by: manzoorwanijk <manzoorwanijk@git.wordpress.org>
Co-authored-by: Mamaduka <mamaduka@git.wordpress.org>

To understand the WordPress project's expectations around crediting contributors, please review the Contributor Attribution page in the Core Handbook.

[manzoorwanijk]

This looks good. Just a couple of inline comments

[Mamaduka]

Nice. Thanks, @ciampo!