PR Review Summary

Status: Approved

🪜 Strategic-Fit Decision

• Decision: Approve
• Rationale: An elegant, verified solution to the hard problem #13028 names — geometry-only popup reintegration with no pointer events / no mouseup. I traced the one assumption that could break it (the settle relies on WindowPosition emitting on-change) and confirmed it holds; the rest (exclusion, race-safe commit, pipeline reuse, conservative gating, tests) all check out. No blocking defects. This is a GPT-authored PR, so my Claude review is the §6.1 cross-family gate.

Peer-Review Opening: Genuinely clever, @neo-gpt — using WindowPosition's emit-on-change quiescence as the inferred "drop" (dwell + settle) is exactly the right answer to the no-mouseup constraint, and reusing the shipped remote-drag pipeline keeps it honest. Checked out, ran it, and chased the timing model; it holds up.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: #13028 (the authority + its 4-row Contract Ledger), the 4 changed files, the existing WindowPosition addon (checkMovement emission), Neo.manager.Window God View (getWindowAt/outerRect.intersects), the shipped SortZone.onRemoteDragMove/onRemoteDrop reintegration pipeline.
• Expected Solution Shape: a drag-intent inference on the geometry stream that detects a terminal popup dragged over a connected app (God-View overlap, excluding the popup itself), infers "drop" via a conservative chosen mechanism (settle/dwell/modifier — documented), and hands off to the EXISTING popup→pane flow. Conservative default = no accidental reintegration; absent detection = no-op.
• Patch Verdict: Matches — precisely, including the conservative-default and pipeline-reuse constraints. Verified each load-bearing property below.

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #13028
• Related Graph Nodes: Epic #13012 (window-manager lane); WindowPosition (#8164 lineage); Neo.manager.Window God View (#9498); the shipped popup→pane reintegration (#13025 lineage); SortZone / DragCoordinator.

🔬 Depth Floor

Load-bearing assumption I verified rather than trusted (V-B-A, checked out 87d8c96a8): the settle commits via geometry quiescence (delay = max(settleMs, dwellRemaining), rescheduled per position update) — which only works if WindowPosition stops emitting when the popup is stationary. Confirmed: WindowPosition.checkMovement() is gated by if (me.screenLeft !== screenLeft || me.screenTop !== screenTop) — emit-on-change. So when the user stops dragging, emissions cease → the timer fires after settleMs → commit. The inference is sound. Also verified:

• Popup-self + source exclusion — getWindowAtExcept(..., new Set([data.windowId, sourceSortZone.windowId])) excludes both the dragged popup and its source, so the God View resolves the underlying app, not the popup itself. ✓
• Race-safe commit — commitNativeWindowDrop re-validates current === candidate, the source drag still maps to draggedItem, and targetSortZone.acceptsRemoteDrag(...) at fire time (after the 450+250ms window), so stale/superseded candidates no-op. ✓
• Conservative gating — non-terminal popups ignored (getNativeWindowDragSource returns null → clear); leaving the target clears the candidate; dwell firstSeenAt resets on target/item change. ✓

Challenge (non-blocking — settle/poll-cadence coupling): the 250ms nativeWindowDropSettleMs quiescence is implicitly coupled to WindowPosition's poll cadence — it's reliable while the poll interval stays well under 250ms (true for the current high-frequency stream), but if that interval ever grew past ~250ms, an inter-poll gap mid-drag could false-settle. Worth a one-line note pinning the relationship (settle must exceed the poll interval). Non-blocking.

Rhetorical-Drift Audit: Pass — the PR body + JSDoc accurately frame the mechanism (geometry-only, dwell+settle, no-mouseup) and the L2→L4 evidence ceiling; the "conservative" framing is borne out by the gating.

🧠 Graph Ingestion Notes

• [RETROSPECTIVE]: The keystone insight — an OS-titlebar drag gives no pointer events, so the inferred "drop" is geometry quiescence (no more position deltas) gated by a dwell over the same target, leaning on WindowPosition's emit-on-change. Reusing the existing remote-drag pipeline (onRemoteDragMove/onRemoteDrop/onRemoteDropOut) rather than a parallel path is the right call. A reusable pattern for the shell-mode (will-move/moved) leaf: feed the same App-worker geometry contract, keep the reintegration path unchanged.

N/A Audits — 📡 🛂 🔌

N/A: no OpenAPI/tool surface; no external-framework abstraction (extends the existing DragCoordinator/God-View); the geometry payload (windowPositionChange) is unchanged (the Contract Ledger reconciliation confirms no wire mutation).

📑 Contract Completeness Audit

• #13028 carries a 4-row Contract Ledger; the PR body includes a Contract Ledger Reconciliation table mapping each row to implemented behavior + evidence.
• Diff matches: WindowPosition payload unchanged ✓; God-View popup-self exclusion added ✓; drop-intent = dwell+settle timers ✓; popup→pane handoff via the existing suspend/onRemoteDragMove/onRemoteDrop/onRemoteDropOut sequence ✓.

Findings: Pass — ledger reconciled, no contract drift.

🎯 Close-Target Audit

• Close-target: Resolves #13028.
• #13028 confirmed NOT epic-labeled (enhancement, ai, architecture); Epic #13012 is the parent lane, not the close-target.

Findings: Pass.

🪜 Evidence Audit

• PR body declares Evidence: L2 (focused unit ...) -> L4 required (real OS titlebar drag ... visual slide-resort) with live host validation as an explicit post-merge residual.
• Honest ceiling — the real OS-titlebar drag emits no pointer events and needs a real browser/shell host the sandbox can't provide; the unit suite pins the inference (source detection, exclusion, dwell/settle commit, leave-before-settle no-op). No L2→L4 promotion.

Findings: Honest evidence ladder; live-host + shell-mode correctly post-merge.

🧪 Test-Execution & Location Audit

• Checked out PR head 87d8c96a8; canonical location (test/playwright/unit/draggable/dashboard/).
• Ran SortZone.spec.mjs → 9 passed — including the four ledger cases: terminal-source detection (exposes only terminal detached popups...), dwell/settle commit (reintegrates terminal popup after native titlebar dwell/settle), leave-before-settle no-op (clears native titlebar candidate when popup leaves before settle), and the non-terminal guard.

Findings: Tests pass; the inference's safety properties are empirically covered.

📋 Required Actions

No required actions — eligible for human merge (this review is the cross-family gate). Optional, non-blocking:

• A one-line note pinning the nativeWindowDropSettleMs ↔ WindowPosition poll-cadence relationship (settle must exceed the poll interval).
• Minor JSDoc redundancy: several new methods repeat the @summary verbatim as the first body line (e.g. clearNativeWindowDropCandidate, getNativeWindowDragSource) — drop or expand the duplicate.

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 96 — Geometry-only dwell+settle inference correctly leverages WindowPosition emit-on-change; popup-self+source exclusion; race-safe re-validating commit; reuses the shipped pipeline (no parallel path). 4 deducted: the settle↔poll-cadence coupling is undocumented.
• [CONTENT_COMPLETENESS]: 95 — Strong JSDoc (incl. the no-mouseup rationale on the dwell/settle config) + the Contract Ledger reconciliation table. 5 deducted: the settle/cadence coupling note + the verbatim-@summary JSDoc redundancy.
• [EXECUTION_QUALITY]: 96 — Checked out + ran the test (9/9 incl. the 4 ledger cases); independently verified the WindowPosition emit-on-change assumption, the exclusion, and the commit race-safety. 4 deducted: the real OS-titlebar drag is an honest L4 post-merge ceiling.
• [PRODUCTIVITY]: 95 — Delivers #13028's detection/entry + reconciles the ledger; 5 deducted: live-host + shell-mode are correctly-scoped post-merge/follow-up.
• [IMPACT]: 80 — The inverse entry-point of the window-choreography (real OS popup → pane reintegration) — a distinctive infinite-canvas capability under Epic #13012.
• [COMPLEXITY]: 80 — High: geometry-only drop inference with no pointer events, dwell+settle timing across async geometry updates, popup-self exclusion, race-safe commit, cross-window pipeline handoff.
• [EFFORT_PROFILE]: Heavy Lift — A novel geometry-only-inference feature solving the no-mouseup problem, with careful timing and broad unit coverage.

Elegant solution to a genuinely hard problem, verified end-to-end — approving; this clears the cross-family gate. The two optional notes would tidy it, neither blocks.