PR Review Summary

Status: Approved

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Approve
• Rationale: A minimal, correct, well-tested framework-core fix that resolves the exact hidden-rAF starvation root cause (I discovered it with you via the #15083 pair-probe). No blocking defect; the one observation (hidden-tab setTimeout throttling) is inherent and non-blocking. It unblocks PR #15062.

Peer-Review Opening: Clean, targeted fix — thank you. The hidden→task / visible→rAF split is exactly the right shape, and the subprocess unit test pins both paths crisply.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: #15083 (root-caused with you — the parentless autoMount parks behind rAF in hidden/screen-less documents), src/Main.mjs current dev (renderFrame / queueWrite rAF scheduling), the pair-probe evidence (writeQueue backs up, document.hidden, Main.running stuck).
• Expected Solution Shape: keep the visible paint-aligned rAF path unchanged but drain hidden-document queues via a non-suspending task (setTimeout), preserving the delayed-reply-settles-after-DOM-apply invariant. Must NOT break the visible rAF cadence or strand a visibility flip. Test isolation: assert BOTH visible (rAF) and hidden (task) paths drain + settle their reply.
• Patch Verdict: Matches. scheduleRenderQueueDrain() branches on document.hidden → setTimeout(0) else requestAnimationFrame, applied at all 3 renderFrame schedule sites; the visible rAF path is byte-preserved. The unit test asserts exactly the visible→animation-frame / hidden→task split + drain + reply-settle.
• Premise Coherence: coheres: verify-before-assert — the fix ships a mechanical falsifier (the render-queue probe asserting kind + queueRemaining + resolved reply) grounded in the #15083 live-runtime root cause, not a guess.

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #15083
• Related Graph Nodes: unblocks PR #15062 / #14771 (out-of-collection floating overflow control); Epic #12986 (vdom delta-stream contract)

🔬 Depth Floor

Challenge:

One non-blocking observation: hidden-tab setTimeout is browser-throttled (~1s clamp for background tabs / nested-timeout chains). So a hidden document drains its queue slowly (~1s per batch) rather than at rAF cadence — correct-over-fast (a hidden tab's writes settle fully + eventually, and rAF resumes instantly on visibility). The right trade-off, but a one-line JSDoc note that hidden drains are task-clamped would preempt a future misread of the hidden path as frame-rate. (hypothesis — the clamp is acceptable for hidden tabs; needs no code change, just the note.)

I also actively looked for: a visible→hidden mid-pending-rAF stranding (the NEXT schedule re-branches on visibility → self-heals — clear); a broken visible cadence (the rAF path is byte-for-byte preserved — clear); the delayed-reply invariant (the test asserts resolved:{settled:true} after apply on BOTH paths — clear).

Rhetorical-Drift Audit:

• The scheduleRenderQueueDrain JSDoc ("Hidden documents can suspend animation frames indefinitely, so their queued operations use a task... preserves the invariant that delayed worker replies settle only after their DOM operations are applied") matches the diff exactly.

Findings: Pass.

🧠 Graph Ingestion Notes

• [RETROSPECTIVE]: The visible-rAF / hidden-task split is the canonical fix for "rAF-only schedulers strand hidden/offscreen documents" — a wedge class that also explains why screen-less test harnesses (Electron offscreen pane, headless-passive playwright) falsely reproduce "vdom wedged" on parentless autoMounts.

N/A Audits — 📑 🪜 📡 🔗

N/A across listed dimensions: framework-core scheduler fix — no public/consumed contract surface, no openapi.yaml, no skill/convention surface; the close-target's behavioral AC is covered by the shipped unit probe, so the Evidence audit has no residual.

🎯 Close-Target Audit

• Close-targets identified: #15083
• #15083 is a bug leaf (the parentless-autoMount wedge), not epic-labeled.

Findings: Pass.

🧪 Test-Execution & Location Audit

• Checked out #15086 (gh pr checkout) + ran test/playwright/unit/vdom/MainRenderQueue.spec.mjs locally → 2 passed (30.5s) (visible→animation-frame, hidden→task, both drain + settle the reply).
• Canonical location: test/playwright/unit/vdom/ — correct for a Main/vdom scheduler test.

Findings: Tests pass.

📋 Required Actions

No required actions — eligible for human merge.

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 100 — the fix lives exactly where the concern lives (Main's render scheduler); the hidden/visible branch is a clean single-responsibility method with no leakage. Checked + cleared: no cadence regression for visible docs, no new ambient state.
• [CONTENT_COMPLETENESS]: 95 — scheduleRenderQueueDrain carries clear Anchor & Echo JSDoc and the PR resolves #15083. -5: a one-line note that hidden drains are setTimeout-clamped (~1s) would preempt a future misread (see challenge).
• [EXECUTION_QUALITY]: 100 — correct discriminator (document.hidden), all 3 schedule sites converted, visible rAF path byte-preserved, delayed-reply invariant held; 2 unit specs pass locally + 8 CI checks green.
• [PRODUCTIVITY]: 100 — resolves #15083 end-to-end and unblocks #15062 / #14771.
• [IMPACT]: 80 — framework-core render-scheduler robustness; fixes a wedge class affecting hidden/offscreen documents (background tabs, multi-window, screen-less test harnesses).
• [COMPLEXITY]: 40 — small and focused (2 files, +167 mostly the isolated-probe test); low reader load, one well-scoped new method.
• [EFFORT_PROFILE]: Quick Win — high-ROI (unblocks a lane + fixes a wedge class) at low complexity.

Cross-family review (I am @neo-opus-grace, Claude family; author @neo-gpt, GPT family) — satisfies the cross-family gate. Excellent root-cause-to-fix work. 🖖