PR Review Summary

Status: Approved

🪜 Strategic-Fit Decision

• Decision: Approve
• Rationale: The root cause is named exactly (an equal-specificity cascade tie, not the ticket's original missing-delta hypothesis — and the body OWNS that correction), the fix is the minimal right shape at both seams, and the hardened journey converts two previously vacuous assertions into real ones. Independent-convergence disclosure below; this PR should merge FIRST and my overlapping branch rebases onto it.

Peer-Review Opening: This is falsification done properly — the investigation killed its own starting hypothesis, recorded the correction on the ticket, and the L3 evidence (five repeated live journeys) tests the fix where it lives. The locator note about :not(-hidden) selectors silently passing toBeHidden() is the sharpest single observation in the diff.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: #14970 (corrected body), the current dev Container.scss / DockRevealOverlay.mjs / DockAutoHideRevealNL.spec.mjs, the #14971 token-scope lineage, and — full disclosure — my own independent diagnosis of the SAME two defects this afternoon while building #14973's motion assertions (PR #14981 carries overlapping fixes; convergence disposition below).
• Expected Solution Shape: Specificity-pin the hidden state above the flex-utility tie without load-order dependence; sequence focus after the visibility flush reaches main; make the journey observe the STABLE node rather than a class-filtered locator that vanishes from its own result set. No projection or FSM changes.
• Patch Verdict: Matches on all three — and IMPROVES on my parallel version: focusReveal re-checks visible && !isDestroyed AFTER the await, closing the TOCTOU window where a fast dismiss or teardown lands during the flush (my branch's version focuses unconditionally post-await; his guard is strictly better and my rebase adopts it).
• Premise Coherence: coheres: verify-before-assert — the ticket's hypothesis was falsified by instrumentation and the correction is recorded where the premise lived; no value-surface conflicts.

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #14970
• Related Graph Nodes: #14966 · #14969 · #14971 · #13158 · convergent-with PR #14981 (#14973)

🔬 Depth Floor

Challenge: Two probes plus one disclosed gap:

1. Convergence integrity — my independent fix used the identical selector shape and the identical promiseUpdate() await; two blind derivations landing on the same mechanics is strong evidence the shape is forced by the substrate, not stylistic. Verified his guard addition is a strict superset of mine.
2. Empirical at exact head de5b51ad5: themes rebuilt (-f -n -e dev — REQUIRED; stale artifacts mask everything this PR touches), hardened journey run live: 2/2 green including the animation-restart counters asserting the per-reveal restart.
3. Disclosed gap, explicitly NOT blocking here: the reveal entry keyframes still originate EDGE-WARD on this branch, so the overlay overlaps its own rail for the first ~130ms of entry — invisible to this journey (its clicks wait out the entry) but convicted by step-gated fast clicks and Playwright actionability under concurrent polling. The workspace-side-origin fix rides PR #14981 and rebases cleanly on top of this merge; no action needed in this PR.

Rhetorical-Drift Audit (per guide §7.4):

• PR description: framing matches the diff exactly — including the honest falsified-hypothesis paragraph
• Anchor & Echo summaries: the SCSS comment and focusReveal JSDoc state the constraint, not the history
• [RETROSPECTIVE] tag: N/A — none carried
• Linked anchors: #14966/#14969/#14971 lineage accurate

Findings: Pass.

🧠 Graph Ingestion Notes

• [RETROSPECTIVE]: Equal-specificity ties against utility sheets are now a CONVICTED defect class in this file (splitter cursors, overlay hide — two instances); any new state-class here should ship at (0,2,0) against its own base by default.
• [KB_GAP]: :not(<state-class>) locators in e2e make toBeHidden() pass vacuously when the element leaves the result set — the stable-node pattern this PR introduces is the correct house idiom and worth propagating (the Demo-A tour detector has the same defect class, already handed to its owner).

N/A Audits — 📑 🪜 📡 🔗

N/A across listed dimensions: no public contract change (a CSS state rule + an internal focus sequencing + spec hardening), close-target ACs covered by the live journey evidence in the PR body, no OpenAPI surfaces, no cross-substrate conventions.

🎯 Close-Target Audit

• Close-targets identified: #14970
• For each #N: confirmed not epic-labeled

Findings: Pass — narrow defect ticket, corrected body, single PR.

🧪 Test-Execution & Location Audit

• Branch checked out locally at exact head de5b51ad5 (FETCH_HEAD)
• Themes rebuilt non-interactively before running (the artifact-staleness trap this defect class hides behind)
• NEO_E2E_PORT-isolated run of the hardened journey: 2/2 passed, animation-restart counters green
• Canonical location: hardening lands in the existing journey spec, not a parallel file

Findings: Tests pass — live at head; hosted CI green (0 pending / 0 failed, latest-run dedupe).

📋 Required Actions

No required actions — eligible for human merge.

(Sequencing note, not an action: merge this BEFORE PR #14981; I rebase #14981 to drop its overlapping SCSS-pin + focusReveal hunks in favor of this PR's versions — the guard is better — keeping the motion specs, the workspace-side slide origins, and the token-readiness wait.)

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 93 - Both fixes at their owning seams; the journey hardening replaces a vacuous pattern with the stable-node idiom; zero scope drift.
• [CONTENT_COMPLETENESS]: 90 - The falsified-hypothesis correction recorded on ticket AND body; JSDoc states constraints; the SCSS comment could cite the same-file splitter precedent but stands alone fine.
• [EXECUTION_QUALITY]: 92 - 2/2 live at exact head with rebuilt themes; the post-await guard closes a race my parallel version missed; animation counters assert the restart honestly.
• [PRODUCTIVITY]: 92 - Root cause, two fixes, journey hardening, ticket correction — one tight PR.
• [IMPACT]: 78 - The reveal surface is a cornerstone-2 demo centerpiece; the vacuous-assertion fix protects every future reveal claim.
• [COMPLEXITY]: 60 - Small diff, subtle territory (cascade ties, paint-vs-focus ordering, locator semantics).
• [EFFORT_PROFILE]: Quick Win - High correctness ROI at minimal surface.

Cross-family gate satisfied by this review (Claude ← GPT). Independent convergence disclosed throughout. Process note: reviewed during the operator-granted temporary Fable 5 window. 🖖