fix(signals): two flights through one memo settle as one unit; lane-dirtied zombies run (#3443, #3444)

[ryansolid]

Fixes #3443. Fixes #3444.

#3443 — a shared memo disagrees with its async inputs during overlapping updates

const a = createMemo(() => delay(1000, inputA()));
const b = createMemo(() => delay(1000, inputB()));
const sum = createMemo(() => a() + b());
// setA(1); await delay(500); setB(1);

Before: 3000: A: 1 alone, 3500: B: 1 | Sum: 2. Traced with the reporter map: setB opens T3 and b's pending propagates onto sum — a memo T2 holds — without recomputing it (its inputs' values are unchanged), so the memo's stamped re-entry never runs and T2 never learns it is waiting on b. sum's reader registers on b in T3 (INV-3 is keyed by transaction); T2's verdict at a's landing sees only its own reporters, commits, and reveals A: 1 beside the committed Sum: 0.

Fix: notifyStatus's dependent walk, on a pending propagation reaching a memo stamped by another live transaction, enters that transaction (initTransition(sub._transition) — a merge, or with nothing active an entry that adopts the ambient batch). This is A15's headline case (a shared derivation of both flights settles as one unit) applied at the one moment it is known. Effects are skipped: an effect entangles nothing by itself (A15 shared-hole corollary, #3407) — the plain-getter shape stays parallel (3000: A: 1 | Sum: 1, 3500: B: 1 | Sum: 2), pinned alongside.

Result: 3500: A: 1 | B: 1 | Sum: 2.

Semantic to note: the write that starts the second flight is held with the first when its async work flows into a held memo — page=1 waits with count=1 while details re-asks, even though Sum reads page plainly. Pinned. The existing #3375 boundary-reset pin moves accordingly: the reset now publishes Sum: 2 once instead of the two-pass Sum: 1 | Sum: 2 its comment described as a drain artifact; every other frame in that pin is unchanged.

#3444 — latest() inside a Show stays stale while removal is pending

When the parking batch is the transaction, flush cancels the zombie recomputes its staged writes queued (cancelZombieRecompute) — a zombie renders mainline until the commit disposes it, and the staged world is one it never displays. A zombie dirtied through the lane channel (REACTIVE_OPTIMISTIC_DIRTY: an override, or the latest() companion the write synced) is the exception: the lane's values are the mainline frame, so the still-visible branch showed latest(count) at 0 beside the same read outside at 1. Such a zombie now runs; its pass runs under the lane and its run lands on the lane's effect queue, so a held lane defers it as it defers every other reader's.

Result: 500: Inside: 1 | Outside: 1, 1500: Inside gone.

Verification

• @solidjs/signals: 173 files / 1804 tests green on top of fix(signals): a memo computes under its own lane posture, never its puller's (#3442) #3445. New tests/overlapping-flights.test.ts (Memoized sum disagrees with its async inputs during overlapping updates #3443 shape, the effect-arm control, the held-second-write consequence) and a latest inside Show stays stale while removal is pending #3444 pin beside 2.0.0-rc nested render effect cleanup run too early #3404 in tests/nested-render-effect-async-cleanup.test.ts.
• Docs: A15 status/pins/mechanism amended for Memoized sum disagrees with its async inputs during overlapping updates #3443 with the consequence stated; INTERNALS bullets for both (§3 Transitions, held-children addendum); rules index regenerated (--check clean). Changeset included.
• Floor 23,365 → 23,419 (+54 B, conscious bump to 23,450); three brotli caps ratcheted 0.05 KB with measurements (+ createStore 15679, isPending/latest 11066, observe tier 15681); every other scenario within cap.

[changeset-bot]

🦋 Changeset detected

Latest commit: aa2cfcc

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 11 packages

Name	Type
@solidjs/signals	Patch
test-integration	Patch
@solidjs/web	Patch
@solidjs/babel-plugin	Patch
@solidjs/compiler	Patch
@solidjs/diagnostics	Patch
@solidjs/element	Patch
@solidjs/h	Patch
@solidjs/html	Patch
solid-js	Patch
@solidjs/universal	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[codspeed]

Merging this PR will degrade performance by 10.67%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

❌ 1 regressed benchmark
✅ 159 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	projection derive: write one NESTED field (reference)	2.5 ms	2.8 ms	-10.67%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing fix/async-memo-overlap (aa2cfcc) with next (75c5113)}

[coveralls]

Coverage Report for CI Build 34938687153

Coverage remained the same at 71.842%

Details

• Coverage remained the same as the base build.
• Patch coverage: No coverable lines changed in this PR.
• No coverage regressions found.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

No coverage regressions found.

---

Coverage Stats

Relevant Lines:	1007
Covered Lines:	772
Line Coverage:	76.66%
Relevant Branches:	790
Covered Branches:	519
Branch Coverage:	65.7%
Branches in Coverage %:	Yes
Coverage Strength:	15.08 hits per line

---

💛 - Coveralls

[ryansolid]

Pushed a cap bump on top of 42632d5: the holds carve-out tips the signals: core floor scenario to 8718 B against next's 8691, 18 B over the 8.70 KB cap (check failed on it). Ratcheted to 8.75 KB with the measurement in the note; the comment's "no bump, measured at 8690" described the stamp-only first cut. All ten scenarios pass locally.

Reviewed the carve-out itself: keying on STATUS_PENDING / a staged _pendingValue rather than the stamp matches the #3334 precedent, and the check on sub runs before the recursive notifyStatus(sub, …) that would mark it, so it reads the memo's pre-propagation state. Locally on this head: signals 174/1895, web 86/783 (args-switch-gate 17/17), solid 27/618, rules-index current.

— Claude via Cursor