[christophersesugh]

Scope

Final verification candidate for all unpublished Throughline changes since signed v0.0.2 (ab62dd0), assessed against merged main at 26d4e81, including player UI, persistent workflow trail, narration, scene/hierarchy, Markdown/MDX and packed integrations.

Verification work in this PR

• Add a read-only Firefox/WebKit browser QA matrix, running the same complete browser suite used by regular Chromium CI. Existing Chromium default and publish workflow remain unchanged.
• Add an independent UI regression asserting controls remain visible, inside the toolbar and non-overlapping across desktop/tablet/mobile layouts, without weakening the existing compact-height assertion.
• Record the actual observed source-branch CI evidence in docs/release-qa-post-v0.0.2.md, with version/commit precision, acceptance coverage, limitations, severity-ranked gaps, manual test procedures and release decision.
• No production behavior, API, trace schema, package exports or publication steps are changed.

Actual recorded evidence

Throughline CI #43212453210 succeeded on exact 26d4e81: 235 unit tests / 33 files and 82 Chromium tests / 14 files, TypeScript, preview, build, license/source audits and packed Node/TypeScript/React SSR/MDX/Vite/Fumadocs/Next/Astro/Markdown consumers.

Release findings (not unverified defect claims)

• P1 — Gate not evidenced: V-10 human comprehension evaluation.
• P1 — Gate not evidenced: real narration/device and assistive-technology testing.
• P1 — Gate not evidenced: actual documentation application integration.
• P2 — Repeatable WebKit test failure: Firefox passed all 83/83 tests including new geometry coverage, while WebKit passed 82/83 and repeated the original one-test failure three times: toolbar.browser.test.tsx:30, height 56px, threshold <55px at 1600×900. QA workflow #43214005524. The added desktop/tablet/mobile control-bounds and non-overlap test passed on WebKit, so the observed issue is the strict height assertion, not demonstrated control clipping. Inspect root cause before changing CSS or assertions.
• P2 — Gate not evidenced: calibrated repeatable performance measurements.
• P1 — Release decision: explicit acceptance of proposed integration/compatibility docs; source-history and rights review.
• P1 — Publication process: new package version, signed tag, staged npm and maintainer approval are separate and not requested or performed.

Overall QA outcome: FAIL — not release-cleared. Original green core CI, Firefox full pass, repeatable red WebKit compact-height check despite passing cross-browser non-overlap geometry; manual and governance gates remain open. This PR creates no certification, approval, merge or deployment. See the report for conditions to re-test and close each gate.

How to verify

Normal PR CI should rerun unit, Chromium, distribution and packed-consumer checks. The additional read-only workflow ran Firefox and WebKit, including a newly added 1600/960/375/320px geometry test; the latter passes in WebKit, but the original 56px toolbar height mismatch still fails. Investigate native control metrics, actual clipping and screenshots, and retest the same acceptance condition before closing the defect. Do not disable or weaken assertions simply to get a green check. Human audio and screen-reader testing remains separate.

Review boundary

Verification-only PR. Please review the findings and the new browser evidence; I will not merge or authorize a release.

[christophersesugh]

Final automated verification snapshot (2026-11-12)

QA PR head: fa38735a11c8f50e1d40dd419b748441842d1b2d, based on the post-v0.0.2 source candidate 26d4e81756e28a6c28ac9f0c4b3dbc2e86b6f8c9.

Job	Actual result
Full CI on PR head #43215373061	PASS — 235 unit tests (33 files); 83 Chromium browser tests (15 files); typecheck, preview/build, source/license/pack audits, independent framework/Markdown consumers all passed
Firefox on PR head #43215372995	PASS — 83/83 browser tests
WebKit on PR head #43215372995	FAIL — 82/83 browser tests. src/react/__tests__/toolbar.browser.test.tsx:30 measured 56px but asserted <55px for wide-toolbar height. The same failure also occurred on the original WebKit matrix and its rerun, and on the interim QA head.
New cross-browser-geometry.browser.test.tsx	PASS in Chromium, Firefox, and WebKit; at 1600, 960, 375 and 320px, toolbar controls stay inside bounds and do not overlap

Triage: This is a repeatable WebKit compact-height assertion failure, not demonstrated clipping or overlap at tested sizes. Inspect native selector metrics and screenshot/actual rendering, distinguish an incorrect pixel-specific expectation from a genuine design regression, and make any production changes only on a separately reviewed fix PR. Keep the failing assertion intact until its intended visual contract is justified and verified.

Release decision: FAIL / NOT RELEASE-CLEARED due to the red WebKit check. Real device/browser voice output, screen readers, participant comprehension (V-10), controlled performance, actual documentation app integration, explicit compatibility/governance decisions, source rights review and final signing/versioning are separately unverified. The linked CI checks were executed by GitHub Actions; no independent local or human device test is claimed. This QA PR is verification-only and remains open for maintainer review.

[christophersesugh]

QA-04 remediation linked (not merged)

I opened #32 as a separate implementation PR, preserving this PR's original verification-only scope. The fix reduces toolbar outer vertical padding while retaining native select sizing and the original strict <55px compact-toolbar assertion. The regression now also checks vertical containment and accepts the empirically observed native chapter-select size range (34px in Chromium/Firefox, 42px in WebKit).

On #32 head 4392eaa806a4b031013c667adaf71fb7630917ed, full standard CI passed (235 unit, 82 Chromium, distribution/consumer checks), and cross-browser QA passed (82 Firefox, 82 WebKit). This is verified on the fix branch, not the existing #31 head.

Next: maintainer reviews/merges #32 first; then refresh/retest #31 against the updated main and update its release QA report with the new assessed candidate and provenance. Do not mark the release cleared: participant comprehension, physical-device speech/assistive tech, the real docs app, controlled performance, design/governance decisions, source rights and publication steps remain open gates. No human or manual device testing is claimed.

[christophersesugh]

Follow-up review of #31 after the WebKit remediation

The separate implementation PR #32 is still open/unmerged, with fix head 4392eaa806a4b031013c667adaf71fb7630917ed. Its standard CI #43217293735 passed (235 unit + 82 Chromium tests and packed/consumer audits); its browser matrix #43217293007 passed 82/82 in Firefox and 82/82 in WebKit. The strict compact-height check remains in place and the updated test also verifies vertical containment. A formal non-approving review comment has been submitted on #32.

This does not change #31's current tested revision: #31 remains at fa38735a11c8f50e1d40dd419b748441842d1b2d; its QA report and workflows reflect the original WebKit failure, not a combined checkout with #32. Please review/merge #32 separately if you accept its scope, then refresh #31 against the resulting main, update the verification report and rerun all checks with its additional cross-browser geometry test (expected total 83 browser tests, not assumed to pass until run). Do not merge #31 as a final release-clearance record with stale failed results.

Status: PR #32's focused WebKit fix has green automation; #31's updated full release verification is not yet completed. Real-device narration, assistive technology, V-10 comprehension, actual docs-app integration, controlled performance, governance, rights, versioning and release authorization remain separate open gates. No merge, approval or publication has been performed.