test: require every e2e fixture build to be reached by a spec

[hivecommons-hive]

Test Improvement

Adds tests/e2e-coverage-build-reach.test.mjs, which requires every e2e
coverage fixture build to be reached by at least one spec.

The fixture-build list is a directory listing, by design: every
tests/e2e/fixtures/data-<name>/ directory declares one more build, and
npm run build:e2e:coverage compiles it into build/e2e-coverage-<name>.
The module header in tests/tools/e2e-data-fixtures.cjs states the cost in the
same breath:

Each build is a full Docusaurus compile, so CI time for the end-to-end
coverage job grows linearly in the number of data-<name>/ directories.

Nothing asserted the other half of that contract.

• tests/e2e-data-fixtures.test.mjs guards the directory → build
  direction. Its comment names the failure it watches for — "a spec asserting
  against a route that 404s" — which is a spec without a build, not a build
  without a spec. It then pins the current names with
  names.includes('variant') / names.includes('no-revision'), which keep
  passing after a build's last spec is deleted.
• tests/e2e-coverage-builds.test.mjs only checks that the compiler loops over
  coverageBuildNames() and derives the right out-dir and base URL. It is
  indifferent to whether the resulting site is ever requested.
• Playwright cannot report it: an unvisited build produces no failing test,
  only a slower job.

A spec has to show two things to count — the base path /e2e-coverage-<name>
(anchored so /e2e-coverage-variant is not credited by a hypothetical
/e2e-coverage-variant-extra) and E2E_COVERAGE_BUILD: '<name>', the pairing
every existing variant spec already uses. Either alone is weak: a spec that
names the base path but never resolves the build's data is asserting against a
route it cannot describe; one that loads the data but never names the base path
never opens the site.

A second test pins that the ordinary coverage build is addressed by the site
root rather than a build name, so the loop is not later "fixed" by adding it to
the name list.

Both current builds (variant, no-revision) pass today — this is a
regression guard, not a report of a broken build.

Verification

• TZ=UTC node --test tests/e2e-coverage-build-reach.test.mjs → 2 passed.
• Mutation check: creating tests/e2e/fixtures/data-orphan/ makes it fail with
  orphan: no spec navigates /e2e-coverage-orphan; the build is a full Docusaurus compile on every run that covers nothing.
• npm run test:unit:coverage:check → exit 0; the new file is itself
  100.00% lines / 100.00% regions, and src files stays 100.00/100.00.
• npx prettier --check → clean.

Coverage evidence

This is a test-infrastructure finding, not a coverage gap: src/ is at
100.00% lines in both lanes.

• Unit: TZ=UTC node tests/tools/coverage-report.mjs at main 3659352 →
  src files 100.00% lines / 100.00% regions (9900/9900, 2873/2873).
• E2E: artifact e2e-coverage (id 11825781781) from CI run
  38418930699,
  head 1e37cdc, 2026-10-13, sealed passed, 204 captured scripts →
  src files 100.00% lines, 97.33% regions (656/674). All 18 residual
  uncovered regions are guard/fallback arms of the documented region-union
  class (see the "What the union credits" note in
  tests/e2e/data-variants.spec.js and issue 1256 ([quality] the e2e region union never credits a guard arm taken only in a sibling fixture build #1256)), not unexercised code.

The two lanes are reported separately rather than merged: they are different
instruments (Node coverage over build-time modules vs. V8 coverage over browser
bundles) with no compatible raw form. A fresh local e2e run at 3659352 was
attempted to replace the 6-day-old artifact, but chromium could not launch in
the analysis environment (libglib-2.0.so.0 missing, no root).

Related Issue

None — this PR carries the finding in full. Two attempts to file it as its own
issue were auto-consolidated by the issue relay into issue 1625 (#1625), which is a
different finding (the unexecuted-chunk arm at
tests/tools/e2e-coverage-report.mjs line 835, already claimed and
stage/040-covered-by-pr). There is no file or behaviour overlap between the
two, so this PR deliberately does not mark that issue resolved. If a maintainer
wants a tracking issue for this, it needs to be opened by hand.

Overlap check

Disjoint from every open hold-gated PR. PR 1626 and PR 1627 touch
tests/tools/e2e-coverage-report.mjs, PR 1627 also e2e-coverage-run.mjs, PR 1629
e2e-coverage-global-setup.mjs; this PR adds one new test file and touches no
existing file. It reads tests/tools/e2e-data-fixtures.cjs and
tests/e2e/*.spec.js, which no open PR modifies.

---

Filed by quality agent (hold-gated mode). Human review required.

— hive: agent=quality backend=copilot model=claude-opus-5 copilot=1.0.88

[hivecommons-hive]

Important

Held for human review by the hive's ACMM level gate.

This PR was opened by the "quality" agent while Hive policy required a human checkpoint for that agent. Agent PRs are held at ACMM L3–L5; an outreach PR that targets a repository this hive does not manage is held at every level because it publishes project-facing communication into someone else's project.

Hive will keep the hold label until a human removes it. Operators can make a deliberate one-off release during an ACMM level change with release_level_holds=true, but level changes never release this hold automatically.

[netlify]

✅ Deploy Preview for endusers ready!

Name	Link
🔨 Latest commit	332ebbb
🔍 Latest deploy log	https://app.netlify.com/projects/endusers/deploys/6ad6234851858356ae76b5e6
😎 Deploy Preview	https://deploy-preview-1636--endusers.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[mrbobbytables]

Priority: P3. This adds an E2E harness reachability guard; no current user-facing failure is reported.

[mrbobbytables]

The fixture-to-spec guard only scans immediate children of tests/e2e via readdirSync(specDir). Playwright collects nested specs recursively, so a valid nested spec could be the only consumer of a fixture build while this test reports it as orphaned. This is the same non-recursive spec-discovery pitfall corrected in #1585/#1586. Please use the repository’s recursive spec discovery pattern and add a nested-spec fixture case before merging.

[mrbobbytables]

Updated this branch in 332ebbb: spec discovery now uses recursive readdirSync paths, and a temp-directory regression test proves a nested .spec.js file is included while a non-spec is excluded. Focused test passes (3/3), Prettier passes, and npm run check:duplication passes. Required CI is running; the PR remains unmerged and its hold label is unchanged.

[mrbobbytables]

The recursive discovery fix and nested-spec regression test address my request. The updated branch passes the focused test, duplication gate, and required CI checks. I am leaving the existing hold label in place; this is not a request to merge.