[arun279]

Removes repository scripts and tests that rebuilt checks pnpm, oxlint, actions/cache, TanStack Query and Reassure already provide. Apart from the cache buster wiring, no product code changes.

What was deleted and what replaces it

Deleted	Replacement
scripts/check-changed-coverage.mjs, scripts/core-coverage-scope.json, the "Gate changed core coverage" CI step, the Coverage-Rationale: PR-body escape, test/ci/changed-coverage.test.ts	Nothing. It only counted core Vitest coverage, so it failed PRs whose core hooks were covered by the native Jest suites. The repo had no global coverage thresholds, and nothing else read the report, so vitest run no longer collects coverage and @vitest/coverage-v8 is removed.
scripts/write-buster.mjs, generated packages/core/src/runtime/persist-buster.ts, buster:check, buster:bump	TanStack's documented buster option, set to the expo-updates updateId (the embedded update's id in a store build, the downloaded update's id after an OTA update). When updates are off, as in development, it falls back to expo-application's version and build number. Every build or update now starts with one cold sync. A new composition-root test checks that a cache saved by an earlier build, or by an earlier update of the same build, is removed and not restored.
scripts/check-node-version.mjs, the preinstall and check:node scripts	engineStrict: true in pnpm-workspace.yaml, plus the existing engines.node. Checked locally: with a mismatched Node, both pnpm install and pnpm run exit 1 with ERR_PNPM_UNSUPPORTED_ENGINE. Without engineStrict they only warn.
scripts/jest-cache-key.mjs, test/ci/jest-cache-key.test.ts	An actions/cache key from hashFiles('.nvmrc', 'pnpm-lock.yaml', 'packages/native/babel.config.js', 'packages/native/jest.config.js'). The repo has no local Babel plugins, so those files cover what the script hashed.
scripts/check-quality-budget.mjs, test/ci/quality-budget.test.ts	oxlint typescript/ban-ts-comment (bans @ts-expect-error, @ts-ignore and @ts-nocheck), and oxlint no-warning-comments with the terms oxlint-disable, eslint-disable and lint/complexity/noExcessiveCognitiveComplexity. That covers Biome complexity suppressions too, so no grep is needed. The script's single-owner checks for DAY_MS, the day-key formatter, storage keys and readFailureOf are dropped, with no replacement.
test/support/stall-guard.ts, its reporter entry, test/ci/stall-guard.test.ts and fixtures, test/ci/ipc-serialization.test.ts	Each job's timeout-minutes.
test/support/git-env.ts, test/ci/git-env.test.ts	The repository Vitest config deletes GIT_* from the environment before workers start. Checked locally: with GIT_DIR pointed at another repository, the git-using tests pass and leave it untouched. Without the strip, they fail.
test/ci/local-jest-workers.test.ts, test/ci/spell-coverage.test.ts, test/ci/executable-scripts.test.ts	Nothing. They checked config text against itself.
The job-timeouts table of slowest run times	A 14-line test that every job with runs-on sets timeout-minutes.
scripts/check-native-assets.mjs, .native-assets.json, test/ci/native-assets.test.ts	Nothing. It was a byte-exact change detector.
scripts/check-render-counts.mjs, scripts/render-count-baselines.json, the base-tree checkout and both base-measuring steps in render-performance, the Render-Count: PR-body escape, .github/actions/use-tree-node, test/ci/render-performance.test.ts, test/ci/second-tree-node.test.ts	Each Reassure scenario now asserts its render count from measureRenders itself (expect(meanCount).toBe(n)), using the same numbers as the old baselines. render-performance is now a plain pnpm --dir packages/native perf:render run. The 20% duration gate on "Library first mount" is dropped. The base-tree action and its test only existed for the base checkout.

Also removed:

• The check job's fetch-depth: 0 and pull-requests: read, and the same two settings on render-performance. Only the deleted gates used them.
• The README.md text about the removed escapes, plus a stale mention of the Binary-Size: escape and the footprint-base job.
• The cspell words only the deleted files used.

The footprint job already only posts a report. No Product-Growth, Binary-Size or Comment-Load parsing remained.

Required checks

.github/required-checks.json is unchanged, because every job keeps its name. No ruleset changes are needed.

Net lines

Category	Before	After	Net
scripts/	1,238	708	-530
test/ (test/ci and test/support)	2,144	1,280	-864
of which test/ci	1,939	1,128	-811
.github/	921	831	-90
Product code (packages/core/src, packages/native/src, packages/native/app)			-1
Native tests (render count assertions, buster test)			+38
Root config, lockfile, README, .native-assets.json			-366
Total			-1,813 (+111, -1,924)

Verification

• pnpm check passes.
• actionlint is clean.
• pnpm --dir packages/native perf:render passes.
• Each new assertion was broken once and failed as expected: a wrong render count, a buster that ignores the update id, a job without timeout-minutes, and a TypeScript or complexity suppression comment.

[github-actions]

Pull request footprint

measurement	base to head
Product code lines in core/src, native/src, native/app, and native/modules	-2
Test lines in test, packages/*/{test,__tests__,e2e}, and .maestro	-744
CI and tooling lines in .github and scripts	-580
Product comment lines identified by a comment prefix	+1

JavaScript bundle at head	size
Expo ios bundle	5624 kB
Expo android bundle	5662 kB