feat(signals,diagnostics): serializable re-run records, artifact timeOrigin (v7), observe idle-cost cap

[ryansolid]

The remainder of observe-tier-plan.md PR B: the attribution engine's records become wire-shaped before the RC stabilises, and the observe tier's idle cost becomes a cap. origin — the third leg of PR B — already landed through the Sentry workstream (withOrigin, CallEvent.origin, currentOrigin()).

RerunEvent is serializable as emitted

• node (a live Computed) is gone. nodeId names the scope: the engine's existing per-node id — the one ChangeOrigin.run and the relay/cycle checks already key on — stable across a scope's runs in the process and distinct between scopes, so runs of unnamed effects still fold to one scope offline.
• OBSERVE.subjectOf(record), which diagnostics already had, now answers for re-run records too. Keyed by the record object, so the node lives exactly as long as some consumer holds the record — the lifetime it had when the record carried it. A copy that left the process has no subject.
• attribution.why(target) / subscriptions(target) unchanged. Only the engine's own recordCosts/why() and two tests read .node; @solidjs/diagnostics was stripping it at three sites and now stores re-runs verbatim (RerunRecord is an alias of RerunEvent).

Clocks: the anchor travels once, not a ts per record

Every at the engine and the runtimes emit is on one clock, performance.now(), and exporters (the spike) already anchor with performance.timeOrigin. A second clock per record would cost bytes on every record and drift against the first. Instead DiagnosticsArtifact.timeOrigin (format v7; also on the JSONL meta line and the browser bridge payload) makes every at in an artifact absolute after the fact — re-runs, holds, records, diagnostic data — and lines a server capture up with the browser session it served. RFC 08 documents the contract.

Idle wiring is a cap

packages/signals/tests/observe-idle-cost.test.ts runs one graph-heavy workload (1000 signal→memo→memo→effect chains, 10 write passes) against the built prod and observe artifacts in one process, interleaved best-of-5, and caps observe/prod at 1.25 — measured 1.03–1.09 with no hooks installed. A round over the cap re-measures (up to three): worker-thread contention is intermittent, a regression is not. Held across four full-suite runs with the web and solid suites running in parallel.

Engine diet — measured, deferred

Ranking every engine function by minified weight: the console face (formatRerun/formatCause/logRerun) is ~0.45 KB gz of the engine's 11.7 KB — 4% — and formatOrigin another 0.15 KB, which the spike calls for span names. The plan's premise is true but small. What a record-consuming production adapter never calls is the in-process query/fold surface (costs, feedback, why, history, ~1.5 KB gz); shedding it means splitting attribution into tree-shakable named exports — an API change for ~2 KB (17%). Written up in the plan's PR B section as a decision rather than taken here.

Verification

• signals 2736 (incl. the new tripwire and a serializability test: no node key, JSON round-trip equal, nodeId stable/distinct, subjectOf returns the node, why() by accessor and by node), diagnostics 34, solid 620 + 825 types, web 986 + 185 types.
• Size caps unchanged (attribution scenario −8 B; wiring floor identical).
• Changeset: @solidjs/signals, @solidjs/diagnostics (patch, prerelease).

Deferred to release, per direction: the vite-plugin work (C5, the SSR componentNames bump).

[changeset-bot]

🦋 Changeset detected

Latest commit: 47601b5

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
@solidjs/diagnostics	Patch
test-integration	Patch
@solidjs/web	Patch
@solidjs/babel-plugin	Patch
@solidjs/compiler	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

[coveralls]

Coverage Report for CI Build 35086846455

Coverage remained the same at 71.46%

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:	1018
Covered Lines:	772
Line Coverage:	75.83%
Relevant Branches:	790
Covered Branches:	520
Branch Coverage:	65.82%
Branches in Coverage %:	Yes
Coverage Strength:	14.94 hits per line

---

💛 - Coveralls

[codspeed]

Merging this PR will improve performance by 16.9%

⚠️ 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 improved benchmark
✅ 174 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	projection derive: write one NESTED field (reference)	2.1 ms	1.8 ms	+16.9%

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing attribution-records (47601b5) with next (8c80a67)}