Content Quality: Concise Briefing (479 words by script, range 300-800) with Overview / What We Know / What We Don't Know structure. Clearly labelled a beta/prerelease in the headline (5.1.0-beta.1), summary ('a prerelease'), overview ('marked as a prerelease') and in What We Don't Know. Per-feature attribution to the correct PR is accurate. Authors of the PRs are not named in the article, so there is no credit to misattribute.
Source Verification: All four sources are first-party github.com pages (vitest-dev/vitest release page and PRs), on the allowlist; no third-party outlets, so sourcing is first-party only, which is appropriate for a release-notes Briefing. Snapshots source-0..3.html.gz (status 200, suspicious_patterns null for all four) were decompressed and text-extracted. source-0 (release page): shows 'Pre-release' badge, released 19 Oct 16:42, tag v5.1.0-beta.1, Features list contains verbatim 'Add --last-failed to run only failed tests' (#11558), 'Introduce --since-last-run flag' (#11516), 'Support Temporal.Instant and Temporal.ZonedDateTime in vi.setSystemTime()' (#11321), 'Expose per-retry/repeat attempt result', 'Support fake timers for node:timers and node:timers/promise'; browser features 'Mock modules through an import map', 'Drop experimental label from trace view', 'Introduce experimental pointer API', 'Capture aria snapshot in trace view'; fixes 'Change default port to 23315' (#11640, listed under Bug Fixes > browser), 'Rerun setup files after --changed filtering', 'Return the mock to concurrent imports'; perf 'Reuse fsModuleCache in static collection'. Live re-fetch via GitHub API (releases/tags/v5.1.0-beta.1): prerelease=true, draft=false, published_at 2026-10-19T16:42:33Z, body matches the snapshot. source-1 (PR #11558, author sheremet-va, merged 2026-10-19): snapshot and API body confirm CLI-only flag, results.json stores full names of failed tests per file, 'A failure outside of tests (hook, collection) reruns the whole file.', names instead of ids 'so adding or moving tests in the file keeps the selection', works with vitest list and other filters, in watch mode only the first run is narrowed. source-2 (PR #11516, sheremet-va, merged): confirms run tests whose dependencies changed since last run, 'doesn't require git and reuses a cached dependency tree instead of parsing every file on each invocation', exposes a resolver object to custom vcs providers used by a new mtime provider that --since-last-run uses under the hood. source-3 (PR #11640, hi-ogawa, merged): confirms previous default 63315 'can be reserved by Windows when Hyper-V, WSL2, or Docker Desktop is running', EACCES vs Vite EADDRINUSE fallback 'so the run fails', 23315-23318 unassigned in IANA registry. All direct quotes verified verbatim; the snapshots were not thin (full PR bodies present), so no fallback was needed beyond the API cross-check. Body URLs equal article.sources (no orphans either direction).
Factual Accuracy: All dates, flag names, semantics and quotes confirmed. --last-failed (reruns failed tests, CLI-only, whole-file rerun for hook/collection failures, watch mode narrows first run only) and --since-last-run (dependency-change based, no git requirement, cached dependency tree, mtime provider) are attributed to the correct PRs (#11558, #11516); the port change to #11640. Temporal scope: the release notes and uncited PR #11321 (read for context only) confirm the new support is for Temporal.Instant and Temporal.ZonedDateTime passed to vi.setSystemTime(); PR #11321 also says vi.useFakeTimers({ now }) already accepted them via Sinon and that a ZonedDateTime sets the instant, not the system time zone. The article's body states the two types and the setSystemTime() scope accurately and makes no claim about Temporal.Now or other Temporal APIs being faked. The article does not cite PR #11321 for the Temporal item; the claim is supported by the cited release notes.
Overall Assessment: APPROVE. Every claim, number, date and quote traces to the four first-party sources; beta status is clearly and repeatedly disclosed; flag semantics and caveats match the PR descriptions; no orphan links; no suspicious_patterns hits. Script verdict of APPROVE stands.