fix(ci): bound the mcp-conformance-fixture beforeAll with hookTimeout

[willgriffin]

Summary

packages/mcp-conformance-fixture/vitest.config.ts set testTimeout: 240_000
but no hookTimeout, so its beforeAll ran against vitest's 10s default —
hooks do not inherit testTimeout. On a loaded runner that hook measured
10,022ms and ejected PR #2230 from the merge queue.

The fix is hookTimeout: 240_000 alongside the existing testTimeout, matching
the repo's prevailing "match testTimeout" convention (smrt-svelte, ads, chat,
cli, content, …; core goes higher at 60s/120s; bundle-gate is the one package
that deliberately diverges).

The diagnosis was verified, not assumed

Everything that could have been supplying a bound was ruled out:

• The hook takes no per-hook timeout argument (beforeAll(async () => {…})).
• vitest.workspace.ts is a re-export shim, not a vitest workspace config —
  it exports smrtVitestPlugin and a setup path. There is no root vitest config.
• smrtVitestPlugin sets no timeouts (the only Timeout hits in
  packages/vitest/src are two setTimeout sleeps in test-db.ts).
• CI passes no timeout flag or env: the package's test script is a bare
  vitest run, invoked through turbo.

Two positive reproductions, both against the unmodified hook:

probe	result
vitest run --hookTimeout=1	Error: Hook timed out in 1ms. — 1 file failed, 2 tests skipped
probe config with hookTimeout: 1	same failure

That is the CI signature exactly (Hook timed out in 10000ms, 1 failed, 2
skipped), and the second probe confirms the config-level option is the
operative one, not just the CLI flag.

Why the hook, and not the tests, is the expensive part

beforeAll does essentially all of the suite's work — slightly more than the
issue described, since it spawns two tsx children, not one:

1. rm -rf + mkdir the generated dir
2. MCPGenerator.generateServer()
3. spawn tsx over stdio → connect Client → negotiate 2026-07-28 → listTools
4. startHttpAdapter() → spawn a second tsx for the HTTP adapter

The two test bodies only issue requests against what the hook already built:
25ms and 192ms locally, against a 240s budget. Locally the hook is ~1.94s of
the 2.16s total; CI's >10s means roughly 5x amplification under load, which is
ordinary for a shard runner spawning child processes.

Blast radius

The fixture is private: true but has a test script, and
scripts/test-packages-shard.mjs excludes only smrt-core — so it runs in the
shared test-packages shard. The 22ms margin was never local to #2230: any PR
in the merge queue could hit it, after a full merge-group validation cycle had
already been spent.

Sibling sweep — none affected

33 of 58 vitest.config.ts files set testTimeout without hookTimeout, but
the omission only matters where a hook does heavy work. Only three packages
have process-spawning or generation code anywhere in their test files, and in
all three it sits in test bodies rather than hooks:

package	hook does	measured	budget
smrt-app-mcp	in-process createHttpServer + listen(0)	conformance test 189ms	15s
smrt-dev-mcp	in-process createHttpServer + listen(0)	conformance test 622ms	30s
tenancy	in-memory sqlite seed (generateServer is inside it() bodies)	—	30s

Their hooks are tens of milliseconds; even 100x CI amplification stays far under
the 10s default. No sibling needs this fix, so the change stays at one package.

Documentation

The issue notes this trap is "worth a line in the testing conventions if it bites
a third time" — it has (smrt-svelte teardown in #1426, then #2220, now this), so
it is recorded in TESTING_STANDARD.md and .claude/rules/testing.md. The
signature is worth naming because it is misleading: the file fails while every
test body reports as skipped, so it does not read as a timeout in the test
you were budgeting for.

Validation

• @happyvertical/smrt-mcp-conformance-fixture full suite: 2 files/2 tests
  passed, 3.26s (transform 376ms, import 845ms, tests 2.33s)
• tsc --noEmit green for the fixture package
• pnpm smrt dev:knowledge-check: fresh, 0 errors, 0 warnings
• pnpm check:agents-chain: 65 chains under the 32,768-byte cap (the two

  80% warnings, packages/users and packages/chat, are pre-existing and
  untouched here)

• biome check on the three touched paths: all three are ignored by
  biome.json (a vitest config and two markdown files), so there is nothing for
  it to lint
• git status clean apart from the three intended files; git diff-tree on the
  commit confirms exactly those three

Closes #2238

{"schema":"hv-agent-run:v1","runtime":"claude","session":"53e2c43c-41e2-4f9f-a63a-996baed10afd","issue":"2238","policy_revision":"1.0.0","status":"complete","validation":["mcp-conformance-fixture full suite: 2 files / 2 tests passed in 3.26s","tsc --noEmit green for the fixture package","diagnosis reproduced twice on the unmodified hook: --hookTimeout=1 and a probe config with hookTimeout:1 both yield the CI signature (1 failed, 2 skipped)","ruled out per-hook argument, root/workspace config, smrtVitestPlugin, and CI flags as alternative sources of the bound","sibling sweep of 33 testTimeout-without-hookTimeout configs: only smrt-app-mcp (189ms/15s), smrt-dev-mcp (622ms/30s), tenancy spawn or generate, and all do so in test bodies, not hooks","pnpm smrt dev:knowledge-check fresh (0 errors, 0 warnings)","pnpm check:agents-chain: 65 chains under cap"]}

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.

[Copilot]

Pull request overview

This PR prevents nondeterministic CI failures in @happyvertical/smrt-mcp-conformance-fixture by explicitly bounding Vitest hook execution time (beforeAll/etc.) with hookTimeout, and documents the repo convention that hook timeouts must be set alongside testTimeout when hooks do meaningful work.

Changes:

• Add hookTimeout: 240_000 to packages/mcp-conformance-fixture/vitest.config.ts to match the existing testTimeout.
• Document the “testTimeout does not apply to hooks” pitfall and its CI failure signature in TESTING_STANDARD.md.
• Add the same timeout guidance to .claude/rules/testing.md for agent/dev workflow consistency.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
packages/mcp-conformance-fixture/vitest.config.ts	Sets hookTimeout to match testTimeout, preventing beforeAll from hitting Vitest’s default hook timeout in CI.
TESTING_STANDARD.md	Adds a testing convention note explaining independent testTimeout vs hookTimeout and the “tests skipped” failure mode.
.claude/rules/testing.md	Captures the timeout rule in the Claude testing rules so it’s less likely to regress.

[willgriffin]

Superseded by #2242, which carries this branch's commits cherry-picked verbatim (same authors, same messages) alongside the other two CI-lane fixes.

Rationale: these three PRs each needed a full validation pass on the very lane they were unclogging, and #2217/#2237 both edit .github/workflows/test-suite.yml. Landing them as one PR at the front of the queue means one validation cycle and one merge, after which the rest of the board can drain.

The merged result is the union of both workflow changes, verified job by job — the three light jobs are hosted per #2236, every heavy job keeps the CI_HOSTED_FALLBACK_ENABLED selector, and coverage-gate keeps #2217's narrowed if: inputs.mode == 'full' (this branch predated that change where applicable). actionlint is clean on all three workflows.

The branch is preserved. Reopen if #2242 needs to be unwound.