[cixzhang]

Maintainer impact

The release-time accessibility contracts can go red on a correct MultiSelector. CreateRow.a11y.chromium.spec.ts fails intermittently in CI (it failed two of the last five release checks) with the polite region reading something other than exactly Create "Urgent". The empty value in the CI log is the region after its auto-clear: the poll had been reading a longer, joined message until then. A red release check on a healthy commit holds the release.

Intended invariant

The contract is unchanged. The live region says 1 result for a partial match, Create "Urgent" when the create row is the whole result (not "No results found"), and Urgent created after Enter. Highlight handling and the created row's selection are untouched.

Change

Since #8244, messages written before the region exposes the previous one are joined, in order, in one update (spec:AST-086 FR9). The test types with pressSequentially, which can send several keystrokes within one frame on a fast runner. The region then reads, for example, Create "Urgen" Create "Urgent": the previous keystroke's message, then the current one. The test compared the whole region text for exact equality, so it passed or failed depending on how many keystrokes fell into a frame.

The three live-region checks now assert the region's newest message (it ends the region text after a space, or is the whole text), which is what the current state is. Test only: no product code and no Changeset.

Evidence

All runs were in Chromium against the Storybook built for the failing commit, with Playwright 1.61.1.

• Failure before the change: 9/20 failures at normal load (4 workers), and 1/40 with the CPU saturated, where slower typing rarely joins. Recording the region showed Create "Urgen" Create "Urgent" in failing runs, then the 2 s auto-clear to "".
• Success after the change: 40/40 at normal load and 40/40 under CPU saturation. The whole file passes repeated runs.
• Product/runtime behavior verified unchanged: no source change. The joined output is the specified FR9 behavior.

Scope

• No intended public API, product behavior, visual direction, or policy change.
• Product changes discovered during the work were removed or split.
• Public text and artifacts contain no internal Meta context.

Testing

• pnpm exec playwright test packages/core/src/MultiSelector/CreateRow.a11y.chromium.spec.ts --project chromium --repeat-each=40 --fully-parallel, at normal load and under CPU saturation, before and after.
• prettier --check and eslint on the file are clean.

Credit: @vjeux wrote this contract in #6829. This change only makes its live-region reads robust to typing faster than a frame.

Posted by Cindy's agent on her behalf.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
astryx	Ready	Preview	Oct 30, 2026 5:40am UTC

[astracat-bot]

AI review status for this pull request.

Review status	Updated
🟢 Reviewed (for maintainers only)	Oct 30, 2026, 6:03 AM UTC

[astracat-bot]

Recommended for approval: the three live-region reads now accept the in-order joined update that spec:AST-086 FR9 specifies, and they still fail on an empty region, a wrong newest message, or a message joined after the expected one. Repeated Chromium runs against a Storybook built from this head reproduce the flake with the old exact comparison and show none with the new reads. Nothing further is needed from you.

[Automated review]