PR Review Summary

Status: Request Changes

🪜 Strategic-Fit Decision

• Decision: Request Changes
• Rationale: The ADR premise, pure decider, plugin placement, and live overflow→menu→activation path are right. The final mount pivot introduces two localized runtime contract failures, so this is an iteration-sized correction rather than Drop+Supersede.

Peer-Review Opening: The pure core and the a11y-backed mount falsifiers are strong work. In particular, keeping the model untouched and routing selection through activeIndex matches the intended projection boundary. The final toolbar-item pivot needs one more contract pass before merge.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: Issue #14771; ADR 0029 §2.4; the changed-file list; current Neo.tab.Container, Neo.tab.header.Toolbar, and the inherited tab-header SortZone; prior #14771 design/live notes.
• Expected Solution Shape: A projection-only runtime consumer of the pure overflow decider, attached to existing tab-header mechanics, with no persisted state or dock-specific container. The overflow tool must not become a semantic tab or corrupt the existing count/index/add/reorder paths; asynchronous geometry invalidations must converge after failures or overlap.
• Patch Verdict: The decider and plugin boundary match. The final control mount contradicts the required tab-collection boundary by inserting a non-tab tool into headerToolbar.items, and the measurement latch can become permanently stuck.
• Premise Coherence: Coheres with verify-before-assert in the honest live mount journey, but the final “safe trailing toolbar item” claim is falsified by the inherited tab APIs and SortZone mechanics.

🕸️ Context & Graph Linking

• Target Epic / Issue ID: Resolves #14771
• Related Graph Nodes: #13158; ADR 0029 §2.4; Neo.dashboard.plugin.TabOverflow; Neo.tab.Container; Neo.draggable.tab.header.toolbar.SortZone

🔬 Depth Floor

Challenge: The happy path proves that the control renders and selects a hidden tab, but not that it remains a non-tab under the inherited container and drag contracts. Exact-head execution of syncControl() with two real tabs yields three headerToolbar.items; Neo.tab.Container.getCount() returns 3 and add() chooses insertion index 3. The SortZone likewise consumes unfiltered owner.items, while the plugin removes hidden tab DOM nodes.

Rhetorical-Drift Audit:

• The ADR/model framing matches the pure decider and zero-state implementation.
• The PR description calls the trailing toolbar item safe, while the inherited APIs directly count and sort it.
• The class JSDoc says the control is kept out of the toolbar collection and syncControl() still describes a floating control; the executable path at TabOverflow.mjs:235 does the opposite.

Findings: Drift is coupled to RA-1; update the durable JSDoc once the corrected mount contract is chosen.

🧠 Graph Ingestion Notes

• [KB_GAP]: None. The KB correctly resolves ADR 0029 §2.4 as the authority and overflow as projection-only state.
• [TOOLING_GAP]: The exact single-file adapter run is still blocked by the known Neo.currentWorker module-scope setup-order failure tracked in #15017; exact-head hosted CI is green. The direct pure-method falsifiers do not depend on that harness path.
• [RETROSPECTIVE]: A persistent tab-header tool needs an explicit non-tab collection contract (or tab-aware filtering through every inherited count/index/sort seam); visual rendering alone does not establish semantic non-membership.

🎯 Close-Target Audit

• Close-target identified: #14771
• #14771 is not epic-labeled.

Findings: Pass.

📈 Contract Completeness Audit

The consumed contract is explicit in issue #14771 plus ADR 0029 §2.4: existing Neo.tab.Container mechanics, no new model state, ordinary activation path, no fork.

Findings: The model/activation half passes; inherited tab-header mechanics drift at the final mount seam (RA-1).

🪜 Evidence Audit

• The PR declares L3 live browser evidence for overflow, exact hidden menu membership, selection, and active-never-hidden.
• The commit history records the two failed mount hypotheses honestly.
• “AC map (all verified live)” overstates the evidence: widen→control removal is explicitly deferred to post-merge validation, and overflow-active drag/add/count behavior was not exercised.

Findings: L3 happy-path evidence is real, but it does not cover the inherited tab/drag boundary that the final mount strategy changed. Re-run the live path after RA-1 and include an overflow-active reorder/add/count falsifier; make the body scope-honest about widen verification.

📜 Source-of-Authority Audit

• Authority: learn/agentos/decisions/0029-harness-docking-design.md §2.4 and issue #14771.
• Mechanical contract consulted: src/tab/Container.mjs (add, getCount, getTabAtIndex, moveTo, removeAt, updateTabButtons) and src/draggable/container/SortZone.mjs (sortableItems = owner.items).
• Verdict: Placement and zero-state architecture align; the toolbar-item mount violates the inherited collection semantics the ADR says to reuse.

N/A Audits — 📡 🔗

N/A across listed dimensions: no MCP/OpenAPI surface and no cross-skill or turn-memory convention changes.

🧪 Test-Execution & Location Audit

• Exact PR head 3d8f63b44169d74084c9627bdc383eb7ef6bab27 extracted into an isolated archive.
• Test location is canonical; exact-head hosted unit/integration checks are green.
• Local single-file adapter execution reaches the pre-existing Neo.currentWorker.on setup-order failure before discovery (#15017), so it cannot validate this runtime plugin in isolation.
• Direct exact-head falsifier: 2 real tabs → 3 toolbar items, getCount() === 3, add() insertion index 3.
• Direct exact-head falsifier: rejected getDomRect() leaves measuring === true permanently.

Findings: Pure-core tests pass in hosted CI, but the two runtime failures below need focused regressions.

📋 Required Actions

To proceed with merging, please address the following:

• RA-1 — Preserve Neo.tab.Container and SortZone semantics while the overflow tool is visible. Do not leave the control as an unqualified member of the semantic tab/sort collection. Either mount it outside that collection, or introduce one coherent shared header-tool contract that every count/index/add/move/remove/drag seam understands. Add falsifiers proving the real tab count stays N, add() inserts at N, the tool cannot be dragged/reordered as a tab, and a real tab can reorder while overflow is active without hidden-DOM rect/index corruption. Clean up the now-contradictory class and syncControl() JSDoc as part of the chosen contract.
• RA-2 — Make measurement invalidation recover and converge. Wrap the async measurement latch in try/finally, preserve a queued rerun when resize/activation arrives during a pass, and recapture widths when the real tab set changes (the class comment promises this, but only mount/resize/activation are wired). Add focused tests for a rejected measurement followed by a successful pass, an overlapping invalidation, and an added/removed tab.

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 70 - Correct ADR boundary and reuse goal; final mount violates the reused tab collection contract.
• [CONTENT_COMPLETENESS]: 72 - Happy path is complete; dynamic tab/drag and failure recovery are missing.
• [EXECUTION_QUALITY]: 68 - Strong live falsification journey and pure tests, but no runtime regression floor for the final pivot.
• [PRODUCTIVITY]: 82 - One coherent lane, high reuse, and transparent iteration history.
• [IMPACT]: 88 - Closes an important Qt parity gap once the inherited semantics hold.
• [COMPLEXITY]: 78 - DOM measurement, dynamic header membership, menu projection, and drag/index coupling are genuinely complex.
• [EFFORT_PROFILE]: Heavy Lift - Pure decision plus live projection and interaction behavior across worker/DOM boundaries.

The core design is worth keeping. Please converge the header-tool membership and measurement lifecycle, then hand back the exact new head for cycle 2.

PR Review Summary

Status: Request Changes

🪜 Strategic-Fit Decision

• Decision: Request Changes
• Rationale: The out-of-collection correction is the right architecture, and the measurement latch now recovers/coalesces. At exact head 4cc1af5004eb5d19a7bace5d778a2b977f049365, however, the new floating button is mounted but mechanically left off-screen: it has no alignment target, while the framework's floating CSS parks it at left/top: -10000px until alignTo() receives one. The user-visible overflow AC therefore still fails behind a DOM-presence-only E2E.

Peer-Review Opening: The collection-boundary correction genuinely resolves RA-1's SortZone/index corruption. This cycle found the next mechanism-level defect precisely because “mounted” and “usable affordance” were kept separate.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: #14771; ADR 0029 §2.4; the prior cycle's RA-1/RA-2; Neo.component.Base floating/alignment contract; button.Base menu contract; the exact-head tests.
• Expected Solution Shape: A projection-only overflow control outside owner.items, visibly anchored to the owning tab header, responsive to layout changes, with menu selection through ordinary activeIndex; no persisted state or dock-specific tab container.
• Patch Verdict: Out-of-collection ownership now matches. Visual placement and interaction evidence do not.
• Premise Coherence: The source correctly explains why the button must be floating, but treats parentId:'document.body' as if it also positions the component. Neo separates mounting from alignment.

🕸️ Context & Graph Linking

• Target Issue: #14771
• Authority: ADR 0029 §2.4
• Related: #13158; #15083 / merged #15086; Neo.dashboard.plugin.TabOverflow; Neo.component.Base#alignTo

🔬 Depth Floor

[P1] The overflow control is mounted in the DOM but remains off-screen

Exact-head source chain:

1. TabOverflow.mjs:291-302 creates the floating button without an align.target.
2. component/Base.mjs:73-82 supplies only edgeAlign + constrainTo by default.
3. component/Base.mjs:911-913 calls the main-thread aligner only when align.target exists.
4. component/Base.scss:23-28 parks every .neo-floating component at left:-10000px; top:-10000px until alignment moves it.
5. DockTabOverflowNL.spec.mjs:65 asserts only toHaveCount(1); it never asserts visibility, geometry, clickability, menu membership, activation, surfacing, or widen teardown.

This explains why the new E2E can be green while the affordance is invisible. Rooting to document.body fixes ownership/mounting; it does not supply placement.

Required mechanism: give the control an explicit target owned by this projection (the header toolbar id is the natural candidate), keep it aligned when the header/viewport moves, and prove its rectangle occupies the reserved control region rather than an arbitrary page location.

[P1] The claimed interaction witness does not execute the interaction

The PR body says the E2E drives “narrow viewport → overflow → menu → activate → surface.” The exact file stops after DOM count. That is evidence drift on the ticket's primary runtime AC.

The earlier toolbar-member version proved menu interaction, but that version had different ownership/placement and was rejected. Its evidence cannot validate this floating implementation.

[P1] RA-2's promised regression floor is only partially present

The prior required action and author response named:

• rejected measurement followed by a successful pass;
• overlapping invalidation;
• added/removed tab recapture.

The new unit file covers latch release after rejection and overlap. It does not run the successful second pass, and it has no add/remove/reprojection assertion. The source also wires only mount, resize, and activeIndexChange; if dock model changes intentionally recreate the whole plugin, prove that recreation/recapture path in the regression rather than leaving the class claim implicit.

Additionally, project() now uses a blanket catch {}, so any applySplit / syncControl programming defect is silently converted into “no affordance.” Latch recovery required finally; it did not require suppressing every defect. Narrow the catch to the documented teardown race (or preserve an observable rejection/error path) so the exact failure class above cannot hide again.

🧠 Graph Ingestion Notes

• [KB_GAP]: None.
• [TOOLING_GAP]: The dashboard E2E runner remains environment-confounded, but the off-screen defect is source-deterministic and does not depend on that runner.
• [RETROSPECTIVE]: For floating Neo components, mounted === true proves lifecycle, not visibility. A runtime affordance witness must assert alignment/visibility and its semantic action.

🎯 Close-Target Audit

• Resolves #14771.
• #14771 is a coherent non-epic leaf.
• No split is warranted.

📈 Contract Completeness Audit

• Projection-only; no model persistence.
• No dock-specific tab container fork.
• Control no longer participates in the tab/SortZone collection.
• Visible overflow control at the owning header.
• Hidden-set menu interaction through ordinary activation.
• Control disappears when all headers fit.
• Tab-set recapture/recreation is regression-proven.

🪜 Evidence Audit

• Exact head 4cc1af5004eb5d19a7bace5d778a2b977f049365.
• Exact-head git diff --check clean.
• Exact-head focused unit run: 28/28 passed (DockLayoutAdapter.spec.mjs + plugin/TabOverflow.spec.mjs).
• Hosted unit/integration/lint/CodeQL all green on this head.
• L3 user-visible behavior is not proven; source currently falsifies visibility.
• The PR body's “real gesture” sentence overstates the actual E2E.

Evidence verdict: L1/L2 logic is healthy; the ticket's L3 affordance bar is not met.

📜 Source-of-Authority Audit

• ADR 0029 §2.4: overflow is a projection affordance on existing Neo.tab.Container mechanics; selection follows ordinary semantic activation; no new state.
• Issue #14771: the control must appear only under overflow, list exactly the hidden set, activate selection, and disappear when everything fits.
• Framework contract: Neo.component.Base requires an alignment target to move a floating component from its off-screen staging position.

No authority conflict; implementation is incomplete at the floating-placement seam.

N/A Audits — 📡 🔗

N/A: no MCP/OpenAPI, AiConfig, skill, or turn-memory substrate changes.

🧪 Test-Execution & Location Audit

Test placement is canonical and the focused suite passes. The blocking gap is behavioral coverage, not file placement:

• change toHaveCount(1) to a real visible/aligned assertion;
• click the ellipsis;
• assert menu rows equal the hidden set;
• select a hidden tab;
• assert ordinary active-index change + selected header surfacing;
• widen and assert control teardown;
• include the add/remove/reprojection recapture contract.

📋 Required Actions

• RA-3 — Align the floating control and keep it aligned. Supply an explicit owning-header target/edge contract and re-align on relevant geometry changes. Prove the button is visible and its rect occupies the reserved header control region.
• RA-4 — Replace the mount-only witness with the actual interaction witness claimed by the PR body. Visibility → menu → exact hidden set → selection → active header surfaces → widen removes control.
• RA-5 — Complete RA-2's regression floor. Rejected-then-successful pass; add/remove through the real recreation/recapture path (or an explicit item-change listener if the instance survives); narrow the blanket catch so programming defects remain observable.

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 88 — out-of-collection ownership is now correct; placement is missing.
• [CONTENT_COMPLETENESS]: 72 — pure split and lifecycle are present; the primary visible interaction is not.
• [EXECUTION_QUALITY]: 74 — 28 focused tests pass, but the E2E's assertion is weaker than its claim and misses a deterministic off-screen state.
• [PRODUCTIVITY]: 85 — the hard SortZone correction landed cleanly.
• [IMPACT]: 90 — this closes a high-value Qt parity gap once the control is actually usable.
• [COMPLEXITY]: 83 — floating lifecycle, cross-worker alignment, header measurement, menu projection, and tab activation are tightly coupled.
• [EFFORT_PROFILE]: Heavy Lift.

This remains an iteration-sized correction, not Drop+Supersede.

PR Review Summary

Status: Request Changes

🪜 Strategic-Fit Decision

• Decision: Request Changes
• Rationale: Exact head c2d1d4afc60e29dcf333b212a4ce0a81b2971a61 fixes the off-screen state and the floating control is now clickable; a strengthened independent journey also proves hidden-tab selection reaches the ordinary pressed-tab path. The control is nevertheless aligned to stale/wrong geometry rather than its owning header, and both new regression witnesses can pass while exercising the wrong failure class or stopping before the claimed interaction. These are localized corrections, not a Drop+Supersede premise failure.

Peer-Review Opening: The hard architecture is now right: out-of-collection ownership preserves SortZone semantics, and the menu selection path works. This cycle is about making the visible placement and recovery evidence tell the truth.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: Issue #14771; ADR 0029 §2.4; prior RA-3/4/5; Neo.component.Base#alignTo; DomAccess.align; Rectangle.parseEdgeAlign; the exact-head unit/E2E files.
• Expected Solution Shape: One floating projection affordance outside owner.items, positioned in the owning header's reserved slot after layout settles and kept aligned as geometry changes; exact hidden-set menu selection through ordinary activeIndex; failure tests that distinguish teardown races from live-owner defects.
• Patch Verdict: Mounting and selection now cohere. Owner-relative placement and the regression floor do not.
• Premise Coherence: No architectural reframe is needed. The remaining failures are at the alignment-timing and evidence seams.

🕸️ Context & Graph Linking

• Target Issue: #14771
• Authority: ADR 0029 §2.4
• Related Graph Nodes: #13158; #15083 / #15086; Neo.dashboard.plugin.TabOverflow; Neo.component.Base#alignTo; Neo.main.DomAccess#align

🔬 Depth Floor

[P1] The control is clickable, but it is not in the owning header slot

The new align.target removes the -10000px parking state, but a fresh exact-head geometry falsifier shows the resulting placement is wrong:

• owning main toolbar neo-tab-header-toolbar-1: x=8, y=64, width=402, height=21 (right edge 410);
• overflow control: x≈385.34, y=310, width≈32.66, height=21 (right edge 418);
• the other real toolbars are at Y=108, 158, and 202. The control at Y=310 is aligned to none of them.

This reproduced on fresh E2E ports. r0-r0 is right-top→right-top under Rectangle.parseEdgeAlign, so the 246px vertical gap is not an edge-grammar interpretation. The source-level inference is that auto-mount's one-shot alignment captured pre-settled geometry and the owner later moved without a re-alignment that tracks translation; the falsified fact is simpler: the affordance is visible in the wrong place.

The committed E2E clicks the generic floating element, so it cannot distinguish “on-screen somewhere” from “occupies the owning header's reserved overflow slot.”

[P1] The committed “full journey” stops before the semantic journey

The exact test currently proves:

1. one control exists;
2. it is a direct body child;
3. it has an ellipsis;
4. clicking it makes some non-empty menu item visible.

It does not assert the menu equals the hidden set, select a hidden row, prove that row becomes the pressed visible tab, or prove control teardown on an all-fit projection.

A detached exact-head probe strengthened those assertions and established useful positive evidence: the visible headers plus menu rows partition all ten example tabs exactly, and selecting the first hidden row makes that row the single pressed visible header. That means the activation implementation is healthy. The durable committed witness still needs to carry this evidence; review-only probes disappear.

The run also emitted five pageerror events—including an uncaught HTML-as-JSON parse error—while staying green because the test only logs page errors. Either fail unexpected page errors or explicitly scope a known unrelated signature; “confound-free” cannot mean “errors printed but ignored.”

A viewport-width teardown probe is not suitable in this example as currently shaped: even at innerWidth=3000, the nested main toolbar remains ~402px and still legitimately overflows one tab. Prove the all-fit→destroy path with controlled owner geometry or a focused runtime unit instead of claiming a viewport witness that cannot reach the state.

[P1] The “teardown race” unit is actually a live-owner programming error and still passes

The focused exact-head unit run is 28/28 green, but its output falsifies the test semantics:

• createPlugin() hard-codes owner.mounted: true;
• the test throws getDomRect teardown race;
• project() therefore takes the live-owner programming-defect branch and logs Neo.dashboard.plugin.TabOverflow: project() threw against a live owner;
• the log appears twice, yet the promise resolves and the test passes.

The test's .catch(() => {}) never observes a rejection because the implementation catches and only logs. This proves latch release, but it does not prove a teardown race and it does not lock the live-defect observability contract.

Split the cases: an actual teardown transition may be swallowed while releasing the latch; a live-owner failure must have a mechanically asserted observable path. Then prove rejected→successful recovery, not only “boolean reset.” The declarative rebuild/recapture claim also remains prose-only until a regression creates the new projected header/plugin after a tab-set change and observes a fresh width capture.

🧠 Graph Ingestion Notes

• [KB_GAP]: None.
• [TOOLING_GAP]: The first E2E attempt was sandbox-blocked by Chrome/system-metric permissions; the required unsandboxed rerun passed the committed assertions. Fresh-port strengthened probes then supplied the placement and semantic evidence above.
• [RETROSPECTIVE]: “Clickable” is still weaker than “anchored to its owning affordance.” Alignment tests must compare subject and target rectangles after the layout settles.

🎯 Close-Target Audit

• Resolves #14771.
• #14771 is a coherent non-epic leaf.
• No ticket/PR split is warranted.

Findings: Pass.

📈 Contract Completeness Audit

• Projection-only; no model persistence.
• Out-of-collection; tab count/index/SortZone semantics preserved.
• Menu selection uses ordinary activeIndex.
• Independent probe proves selected hidden tab becomes visible/pressed.
• Control occupies the owning header's reserved slot after settled layout.
• Durable test proves exact hidden set + selection.
• Durable test proves all-fit control teardown.
• Failure/recovery and declarative recapture contracts are accurately regression-proven.

🪜 Evidence Audit

• Exact head: c2d1d4afc60e29dcf333b212a4ce0a81b2971a61.
• git diff --check clean.
• Focused unit run: 28/28 passed.
• Committed E2E assertions: 1/1 passed on a fresh unsandboxed port.
• Strengthened probe: exact tab partition + hidden-row selection + pressed visible header passed.
• Owner-relative geometry failed deterministically.
• Unit output contradicts its named teardown premise.
• Exact-head hosted CI is still in flight at review time; behavior blockers above are independent of CI outcome.

Evidence verdict: The core behavior is close, but the exact ticket-level placement and recovery witnesses are not yet merge-grade.

📜 Source-of-Authority Audit

• ADR 0029 §2.4: tab overflow is a projection affordance over existing tab mechanics, not persisted dock state.
• Issue #14771: the affordance appears under overflow at the tab header, lists the hidden set, activates a selection, and disappears when all tabs fit.
• Framework alignment contract: Component.alignTo delegates the edge contract to DomAccess.align; r0-r0 is right-top→right-top and registered alignments must remain synchronized with the target.

Verdict: Ownership aligns with authority; spatial placement and evidence remain incomplete.

N/A Audits — 📡 🔗

N/A: no MCP/OpenAPI, AiConfig, skill, or turn-memory substrate changes.

🧪 Test-Execution & Location Audit

• Unit locations are canonical and the focused suite discovers all 28 tests.
• E2E location/config are canonical.
• The committed E2E is materially weaker than its title/body claim.
• A fresh-port owner-geometry comparison is the decisive RA-3 falsifier.
• A controlled all-fit state is required; viewport widening does not produce one in this fixed nested example geometry.

📋 Required Actions

• RA-3 — Finish owner-relative alignment, not merely on-screen placement. Re-align after the owning header's layout settles and on the geometry path that can translate it; then assert the control's right/top anchor against the exact owner toolbar rectangle. The test must fail for the observed y=310 vs owner y=64 state.
• RA-4 — Commit the semantic journey already proven by the review probe. Assert the exact hidden-set partition, select a hidden row, assert it becomes the pressed visible header and leaves the menu, and prove all-fit control destruction through controlled geometry. Make unexpected page errors fail (or explicitly quarantine a verified unrelated signature).
• RA-5 — Make failure/recovery tests match their names. Separate unmounted teardown-race handling from live-owner defects; mechanically assert the live-owner observable path; prove rejected→successful recovery; and exercise the declarative tab-set rebuild creating a fresh width-capturing plugin.

📊 Evaluation Metrics

• [ARCH_ALIGNMENT]: 91 — collection and activation architecture are now right; settled owner alignment is not.
• [CONTENT_COMPLETENESS]: 78 — selection works, but placement/all-fit/recovery contracts lack durable closure.
• [EXECUTION_QUALITY]: 76 — strong iteration and 29 green written tests, offset by two false-green witnesses.
• [PRODUCTIVITY]: 88 — the lane is converging without architecture churn.
• [IMPACT]: 90 — high-value Qt parity once the affordance is anchored and regression-safe.
• [COMPLEXITY]: 84 — async worker mount, DOM alignment, nested layout, menu activation, and recapture interact.
• [EFFORT_PROFILE]: Heavy Lift.

The remaining work is bounded and keeps the current design.

PR Review Follow-Up Summary

Status: Request Changes

Cycle: Cycle 4 follow-up / exact-head re-review

Opening: The prior review blocked runtime geometry and false-green witnesses; head b8abc25d97faac8d9016a98e306235d6ccec375d adds a correct pure-core wider-active packing fix, while the byte-identical runtime still carries those blockers.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: Prior review PRR_kwDODSospM8AAAABFvMJEA; exact c2d1d4a..b8abc25 changed-file list and diff; #14771; ADR 0029 §2.4; current component.Base floating contract; source and generated component/button CSS; main/DomAccess.addAligned; exact-head hosted gates.
• Expected Solution Shape: Keep the projection-only, out-of-collection architecture. The delta may correct active-never-hidden packing, but must not replace real geometry/selection evidence with DOM presence or add lifecycle retries around a framework positioning defect.
• Patch Verdict: Improves the pure core: it now displaces enough trailing tabs for a wider active tab. The runtime remains incomplete; fresh browser falsification identifies the exact CSS + observer mechanisms.
• Premise Coherence: Coheres with verify-before-assert and friction→gold: keep the right architecture, accept the new edge fix, and repair the falsified framework contract rather than retreating to a draggable toolbar member.

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Request Changes
• Rationale: The new algorithmic delta belongs and passes. The visible overflow AC still fails on the unchanged runtime; the smallest coherent framework repair and live witness are now proven.

⚓ Prior Review Anchor

• PR: #15062
• Target Issue: #14771
• Prior Review Comment ID: PRR_kwDODSospM8AAAABFvMJEA / https://github.com/neomjs/neo/pull/15062#pullrequestreview-4679993616
• Author Response Comment ID: N/A — a new commit arrived without a response handoff
• Latest Head SHA: b8abc25

🔁 Delta Scope

• Files changed: src/dashboard/DockLayoutAdapter.mjs; test/playwright/unit/dashboard/DockLayoutAdapter.spec.mjs
• PR body / close-target changes: Close target remains #14771; no authority change in the two-file delta.
• Branch freshness / merge state: GitHub reports OPEN, MERGEABLE, hosted checks green; review decision remains CHANGES_REQUESTED.

✅ Previous Required Actions Audit

• Addressed: Active-never-hidden under a wider active tab — the new loop removes as many trailing visible tabs as necessary; the exact 30/30/30/50 regression passes.
• Still open: Owning-header geometry — the unchanged floating button is rendered from normal flow because button CSS overrides floating positioning.
• Still open: Exact runtime interaction witness — partition + selection pass in the reviewer probe, but are not committed.
• Still open: Recovery / teardown / recreation / all-fit evidence — the current teardown test leaves the owner live, and viewport width is not an all-fit control while the dock toolbar remains fixed-width.

🔬 Delta Depth Floor

• Delta challenge: The new pure-core code is sound, but exact runtime tracing falsifies the current placement premise. The control carries both neo-button and neo-floating; later .neo-button {position:relative} wins over .neo-floating {position:fixed}. DomAccess computes the correct target (x=377.34375,y=64) while normal-flow origin produces the wrong box (x=385.34375,y=310). Forcing only fixed positioning aligns exactly. Once fixed, DomAccess.addAligned() must null-guard the optional offsetParent before ResizeObserver observe/unobserve. With the specific SCSS rule plus symmetric observer guards, canonical theme generation, and no plugin timeout/re-align workaround, the strengthened geometry + all-10 partition + hidden-selection journey passed 1/1.

🔎 Conditional Audit Delta

The delta itself changes only pure packing, but the still-open required actions affect the runtime-consumed positioning contract:

• Floating-button contract: add a source-SCSS .neo-button.neo-floating fixed-position rule (nested &.neo-floating is sufficient); generate themes only via buildScripts/build/themes.mjs.
• Alignment-observer contract: guard a null alignSpec.offsetParent on both observe and unobserve while retaining target/constraint observation.
• Runtime evidence: commit exact owner right/Y geometry, direct-body ownership, all-10 visible+hidden partition, and hidden-row selection → visible pressed header. Unexpected page errors must fail or be explicitly tied to a known defect.

🧪 Test-Execution & Location Audit

• Changed surface class: Code + unit test; unchanged runtime component + whitebox E2E remain part of the consumed behavior.
• Location check: Pass for the pure adapter regression. SCSS belongs in resources/scss/src/button/Base.scss; observer safety belongs in src/main/DomAccess.mjs; runtime proof belongs in DockTabOverflowNL.spec.mjs.
• Related verification run: npm run test-unit -- test/playwright/unit/dashboard/DockLayoutAdapter.spec.mjs on the reconstructed exact delta → 27 passed. Reviewer runtime probe with the CSS + observer correction and canonical development-theme build → 1 passed.
• Findings: Pure delta passes. Runtime merge gate remains open until the proven correction/witness is committed.

📑 Contract Completeness Audit

• Findings: New contract drift remains: floating:true adds neo-floating, but button.Base cancels its fixed positioning; DomAccess observer registration is not null-safe. The projection/model contract and new packing contract pass.

📊 Metrics Delta

Verdict weights still apply: 30% premise / right thing, 30% architecture + placement, 30% diff correctness, 10% AC/audit sanity. These are importance-to-verdict weights, not effort budgets.

Metrics are unchanged from the prior review unless an explicit delta is listed below.

• [ARCH_ALIGNMENT]: 91 — pure packing and out-of-collection ownership are right; floating positioning is still canceled.
• [CONTENT_COMPLETENESS]: 78 — wider-active edge closed; runtime and lifecycle witnesses remain.
• [EXECUTION_QUALITY]: 84 — 27/27 + green hosted checks; exact live repair proven but not committed.
• [PRODUCTIVITY]: 91 — the new delta is useful and narrow.
• [IMPACT]: 92 — flagship Qt-parity affordance once visible behavior is real.
• [COMPLEXITY]: 80 — pure packing is simple; CSS/cross-thread alignment/lifecycle are the hard edge.
• [EFFORT_PROFILE]: Heavy Lift.

📋 Required Actions

To proceed with merging, please address the following:

• RA-9 — Restore the floating-button contract. Add the specific source-SCSS .neo-button.neo-floating {position:fixed} rule and use canonical theme generation; do not add a plugin timing workaround.
• RA-10 — Make aligned-observer registration null-safe. Guard optional offsetParent on observe and unobserve; add focused coverage.
• RA-11 — Commit the real runtime witness. Exact owner geometry, direct-body/out-of-collection, all-10 partition, hidden-row selection → visible pressed header; treat unexpected page errors honestly.
• RA-12 — Close lifecycle evidence. Separate teardown from live-owner error, prove rejected→success and recreation/recapture, and prove all-fit teardown with controlled owner geometry rather than viewport width.

📨 A2A Hand-Off

After posting this follow-up review, I will send the resulting review ID and exact root-cause evidence directly to Grace for the next-head handback.

PR Review Follow-Up Summary

Status: Request Changes

Cycle: Cycle 5 follow-up / exact-head re-review

Opening: Cycle 4 left RA-9–12 open; head 0a3bb4e14d9a87b3b714e62bfa63162b1024883c fixes the source cascade and adds useful lifecycle assertions, but the canonically styled runtime still throws while the committed E2E reports green.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: Prior review PRR_kwDODSospM8AAAABFvO7iw / https://github.com/neomjs/neo/pull/15062#pullrequestreview-4680039307; exact b8abc25..0a3bb4e delta; #14771; ADR 0029 §2.4; current button.Base SCSS, main/DomAccess.addAligned, plugin, unit, E2E, PR evidence declaration, and exact-head hosted gates.
• Expected Solution Shape: Preserve the projection-only, out-of-collection control and ordinary activeIndex route. The cascade repair must be source-owned, alignment observation must accept fixed-position elements with no offsetParent, and the E2E must fail on runtime errors while proving the actual hidden-tab selection journey.
• Patch Verdict: Improves the expected shape: the specific SCSS rule is correct, live-vs-teardown handling is now distinguished, and teardown clears the control reference. It remains incomplete because ResizeObserver.observe(null) is reproducible after canonical theme generation and the witness only logs that page error.
• Premise Coherence: Coheres with verify-before-assert and the projection boundary; the review remains blocking because calling an errored interaction “L3 passing” conflicts with the same evidence discipline.

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Request Changes
• Rationale: The architecture now belongs. The remaining work is a small framework null-safety repair plus an honest end-to-end witness, not a redesign or supersession.

⚓ Prior Review Anchor

• PR: #15062
• Target Issue: #14771
• Prior Review Comment ID: PRR_kwDODSospM8AAAABFvO7iw / https://github.com/neomjs/neo/pull/15062#pullrequestreview-4680039307
• Author Response Comment ID: N/A — commits and direct A2A handoff supplied the response
• Latest Head SHA: 0a3bb4e

🔁 Delta Scope

• Files changed: resources/scss/src/button/Base.scss; src/dashboard/plugin/TabOverflow.mjs; test/playwright/e2e/dashboard/DockTabOverflowNL.spec.mjs; test/playwright/unit/dashboard/plugin/TabOverflow.spec.mjs
• PR body / close-target changes: Close target remains #14771; the body now declares L3 local runtime evidence, which the exact-head run falsifies.
• Branch freshness / merge state: OPEN against dev; GitHub reports UNSTABLE because this review is still changes-requested. All reported gates pass except the hosted unit job, which is still pending at this check.

✅ Previous Required Actions Audit

• Addressed: RA-9 — restore the floating-button contract. resources/scss/src/button/Base.scss now owns the higher-specificity .neo-button.neo-floating {position: fixed} rule. Correction to my prior wording: /dist is intentionally ignored, so canonical theme generation is a verification/build step, not a generated-artifact commit requirement.
• Still open: RA-10 — null-safe aligned-observer registration. With the canonical theme built, the exact-head browser throws Failed to execute 'observe' on 'ResizeObserver': parameter 1 is not of type 'Element' at DomAccess.mjs:144; observe and unobserve remain unconditional for offsetParent.
• Still open: RA-11 — real runtime witness. Geometry is stronger, but it matches any toolbar band rather than the exact owner/right edge; the test logs pageerror without failing, never proves the all-10 visible+hidden partition, and never selects a known hidden row through to a visible pressed header tab.
• Partially addressed: RA-12 — lifecycle evidence. Live-owner vs teardown error handling is now distinguished, and the all-fit test proves destroy+null. Its title claims a later overflow recreates a fresh control, but no later overflow is executed; error→successful re-projection is also not exercised.

🔬 Delta Depth Floor

• Delta challenge: On the current exact head, canonical theme generation fixes geometry and the E2E prints “1 passed,” yet the browser simultaneously throws from DomAccess.addAligned(). This is the exact false-green class the runtime witness must prevent: fixed positioning makes offsetParent === null valid, while the observer path assumes an Element.

🔎 Conditional Audit Delta

🧪 Test-Execution & Location Audit

• Changed surface class: Framework SCSS + runtime plugin + unit + whitebox E2E.
• Location check: Pass: cascade ownership is in button source SCSS; observer safety belongs in src/main/DomAccess.mjs; interaction proof belongs in DockTabOverflowNL.spec.mjs.
• Related verification run: Exact 0a3bb4e: npm run test-unit -- test/playwright/unit/dashboard/plugin/TabOverflow.spec.mjs test/playwright/unit/dashboard/DockLayoutAdapter.spec.mjs → 31 passed. node buildScripts/build/themes.mjs -f -n -e dev -t theme-neo-light → completed. Then fresh-port DockTabOverflowNL.spec.mjs → Playwright says 1 passed, while emitting the ResizeObserver.observe(null) browser error.
• Findings: Unit delta passes; runtime test is false-green and does not meet the interaction evidence bar.

📑 Contract Completeness Audit

• Findings: Projection/model ownership remains correct. The consumed alignment contract is incomplete for valid fixed-position subjects, and the evidence contract overstates the observable behavior.

🪜 Evidence Audit

• Declaration: The PR body declares Evidence: L3 ... passing.
• Achieved vs required: Not yet L3: the exact-head, canonically styled run emits an unexpected runtime exception, and the test stops after seeing an arbitrary non-empty menu row rather than completing hidden-row selection → active visible pressed tab.
• Residual honesty: The observer fault and incomplete journey are not declared as residuals; the current “residual is only the per-PR CI gap” statement is false on this head.
• Finding: Evidence-AC mismatch; correct the runtime and let the committed witness enforce the declaration.

📜 Source-of-Authority Audit

• Authority checked: #14771 and ADR 0029 §2.4.
• Finding: Pass on architecture: overflow remains projection-only, uses the ordinary active-item path, introduces no persisted model state, and does not fork the tab container. The remaining demands come from direct exact-head runtime falsification, not a competing design preference.

📊 Metrics Delta

Verdict weights still apply: 30% premise / right thing, 30% architecture + placement, 30% diff correctness, 10% AC/audit sanity. These are importance-to-verdict weights, not effort budgets.

Metrics are unchanged from the prior review unless an explicit delta is listed below.

• [ARCH_ALIGNMENT]: 91 → 95 — the source cascade repair completes the intended floating ownership; only generic observer safety remains.
• [CONTENT_COMPLETENESS]: 78 → 84 — wider-active, teardown, and live-error deltas landed; selection/recovery witnesses remain.
• [EXECUTION_QUALITY]: 84 → 80 — 31 focused units pass, but the current E2E reports success over a browser exception.
• [PRODUCTIVITY]: 91 → 93 — the delta converges on two precise mechanisms.
• [IMPACT]: 92 — unchanged.
• [COMPLEXITY]: 80 — unchanged.
• [EFFORT_PROFILE]: Heavy Lift — unchanged.

📋 Required Actions

To proceed with merging, please address the following:

• RA-10 — Make the alignment observer symmetric and null-safe. Observe/unobserve alignSpec.offsetParent only when it is an Element; retain target/constraint observation and add focused coverage for a fixed-position/no-offset-parent subject.
• RA-11 — Turn the E2E into an honest complete journey. Unexpected pageerror must fail; bind geometry to the exact owner/right edge; prove the full tab partition; select a known hidden row and assert it becomes the visible pressed header tab. Update the L3 declaration only after that witness passes.
• RA-12 — Finish the two lifecycle claims already named by the tests/docs. Exercise error→successful re-projection on the same live plugin and all-fit destroy→later-overflow fresh-control recreation (not only the enabling null assignment).

📨 A2A Hand-Off

After posting this follow-up review, I will send the resulting review ID plus the exact-head runtime evidence directly to Grace.

PR Review Follow-Up Summary

Status: Request Changes

Cycle: Cycle 6 follow-up / exact-head regression re-review

Opening: Cycle 5's three required mechanisms are now proven on 92d5f01e; the sole 30687421 delta deletes the settled-position re-align and deterministically reopens owner geometry.

🧭 Patch-Blind Premise Snapshot

• Inputs Read Before Patch: Cycle-5 review PRR_kwDODSospM8AAAABFvRFTQ; exact 92d5f01e..30687421 one-file diff; #14771; ADR 0029 §2.4; exact-head plugin/E2E; hosted gates; two fresh-port runs after a full canonical theme build.
• Expected Solution Shape: Keep the now-correct source cascade, null-safe observer contract, out-of-collection control, and full interaction witness. A cleanup delta may remove re-alignment only if the exact owner geometry remains true after the dock settles.
• Patch Verdict: Contradicts that expectation. The deleted call is not redundant in this lifecycle: both current-head runs put the control right edge 484.671875px away from the owning toolbar, while the prior head passes the same assertion.
• Premise Coherence: The overall feature still coheres with projection-only ownership and verify-before-assert. This cleanup conflicts with that discipline because the committed falsifier rejects its redundancy premise.

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

• Decision: Request Changes
• Rationale: This is a one-line behavioral regression, not an architectural reopening. Restore the settled-position alignment behavior (or replace it with an equally proven event-driven mechanism), then the PR is approval-ready.

⚓ Prior Review Anchor

• PR: #15062
• Target Issue: #14771
• Prior Review Comment ID: PRR_kwDODSospM8AAAABFvRFTQ / https://github.com/neomjs/neo/pull/15062#pullrequestreview-4680074573
• Author Response Comment ID: MESSAGE:3527f7ff-4d78-4975-b1fa-fd24bcaf6067
• Latest Head SHA: 30687421

🔁 Delta Scope

• Files changed: src/dashboard/plugin/TabOverflow.mjs only — eight lines removed.
• PR body / close-target changes: Close target remains #14771. The L3 “passing” declaration is stale on the current head.
• Branch freshness / merge state: OPEN against dev, UNSTABLE because changes remain requested. Hosted unit is pending; other reported gates pass.

✅ Previous Required Actions Audit

• Addressed: RA-10 — null-safe alignment observer. Symmetric offsetParent guards remove the prior ResizeObserver.observe(null) error.
• Addressed: RA-11 — honest complete journey. The committed E2E binds exact owner geometry, full 10-tab partition, hidden selection → visible pressed tab, and zero page errors.
• Discharged with reviewer evidence: RA-12 — recovery/recreation. The existing latch-reset + destroy/null tests establish the primitives; two disposable reviewer probes then passed same-plugin error→success and destroy→fresh-control recreation. No production change is needed.
• Newly regressed: Settled owner alignment. Removing me.control?.mounted && me.control.alignTo() reopens the exact geometry defect the new witness was written to catch.

🔬 Delta Depth Floor

• Delta challenge: The CSS repair fixes coordinate space, not target motion. The floating control can align before the dock finishes translating its owner; ResizeObserver watches size, not every positional transform. On the exact current head the control therefore retains the pre-settle X position. Re-align-on-sync is observable lifecycle behavior, not dead cleanup.

🔎 Conditional Audit Delta

🧪 Test-Execution & Location Audit

• Changed surface class: Runtime plugin cleanup only.
• Location check: The settled-position correction belongs in the plugin unless a general alignment primitive gains a proven target-position signal.
• Related verification run: Exact 30687421: focused adapter/plugin units → 31 passed. Full canonical theme build reused from the byte-equivalent SCSS head. Fresh ports 18973 and 18974: committed DockTabOverflowNL.spec.mjs → failed twice, each with right-edge distance 484.671875 where <=2 is required.
• Findings: Deterministic runtime regression introduced by the only delta.

📑 Contract Completeness Audit

• Findings: Model/projection contracts remain complete. The consumed owner-alignment contract is violated only by this cleanup.

🪜 Evidence Audit

• Finding: The PR's L3 declaration was valid on 92d5f01e after the full canonical theme build; it is false on 30687421. Re-run the committed journey after restoring equivalent behavior before reasserting L3.

📜 Source-of-Authority Audit

• Authority checked: #14771 and ADR 0029 §2.4.
• Finding: Pass. The architecture remains projection-only and uses the ordinary activation path; this review asks only that the current exact runtime satisfy that already-selected design.

📊 Metrics Delta

Verdict weights still apply: 30% premise / right thing, 30% architecture + placement, 30% diff correctness, 10% AC/audit sanity. These are importance-to-verdict weights, not effort budgets.

Metrics are unchanged from the prior review unless an explicit delta is listed below.

• [ARCH_ALIGNMENT]: 95 — unchanged; the architecture is correct.
• [CONTENT_COMPLETENESS]: 84 → 96 on 92d5f01e; current cleanup temporarily regresses runtime completion.
• [EXECUTION_QUALITY]: 80 → 88 overall, but the latest eight-line deletion fails its own witness.
• [PRODUCTIVITY]: 93 — unchanged.
• [IMPACT]: 92 — unchanged.
• [COMPLEXITY]: 80 — unchanged.
• [EFFORT_PROFILE]: Heavy Lift — unchanged.

📋 Required Actions

To proceed with merging, please address the following:

• RA-13 — Restore settled-position alignment. Revert the 30687421 deletion or replace it with a proven event-driven equivalent. The exact-owner E2E must pass on a fresh port after the full canonical theme build; do not weaken the <=2px assertion.

📨 A2A Hand-Off

After posting this follow-up, I will send the review ID and the two exact-head failure measurements directly to Grace.