fix(signals,solid): reserve _parent from mangling (two prod/observe cross-package bugs); retire the wall-clock tripwires; PII surface in RFC 08

[ryansolid]

Three of the four remaining in-repo items from the Sentry/observe plan: the observe-tier mangling audit (which found two real bugs), the flaky timing tripwires, and Solid's half of the PII question. The engine diet stays a separate decision.

_parent is the second cross-package owner field

Signals' prod and observe artifacts mangle every _-prefixed property except a reserved list (scripts/mangle-props.mjs); the dev artifact is unmangled, and every suite runs against dev. Two reads of _parent cross the package boundary and only ever worked in dev:

• solid-js client hydration (hydration.ts, markTopLevelSnapshotScope) walks owner._parent to the root to mark the hydration snapshot scope. In the built prod and observe artifacts owner._parent is undefined, so the walk marked the current owner — computations created outside that owner's subtree during hydration read live values instead of the server snapshot (CONFIG_IN_SNAPSHOT_SCOPE never set on them). Prod-only, invisible to the suite.
• The core's owner walks over server owners — ownerPath and OBSERVE.exclude/isExcluded. ownerPath had a server-side shim (located() in solid/src/server/diagnostics.ts, removed here); OBSERVE.exclude on a server owner was a silent no-op outside dev.

_parent joins _name on the reserved list. Cost is mangling only — nothing added to any code path: core floor +8, createStore +20, isPending +32, simple app +20, hydrating +43, hydrating+stores +19, CSR +41, observe CSR −38, attribution −2, frames 0 (bytes, brotli). Two caps bumped (hydrating 19.80→19.85, CSR 15.15→15.20) with notes.

Pinned from both ends:

• packages/solid/test/cross-package-fields.spec.ts — the reserved fields survive in the mangled signals artifacts (and _firstChild does not, proving the mangler ran); every built client artifact of solid-js, @solidjs/web, @solidjs/universal is scanned for ._field accesses, and any signals _ field (parsed from types.ts) not on the reserved list fails with the file and context. Un-reserving _parent makes it flag hydration.ts's walk.
• packages/web/test/server/server-owner-walks.spec.tsx — ownerPath (component labels included) and OBSERVE.exclude over server owners, against the built observe and development artifacts in a child Node under --conditions. Rebuilt without the reservation, both fail on the observe artifact.

Wall-clock tripwires retired

Four timing assertions were costing CI reruns under turbo's parallel load:

• heap-mark-incremental: re-measures a round over the cap (up to three) — the quadratic regime is over the cap every round, a GC pause on the one large sample is not.
• attribution-navigation: asserted settledMs >= 10 after a setTimeout(10), which can fire a hair under 10 ms on the performance.now() clock; now asserts against the wait actually taken.
• lazy-shell-gating / dynamic-shell-gating: proved ORDER (shell before/after a source settled) with wall-clock deltas against timers (shellAt < DATA). Rewritten on shell-gating-harness.ts: every async source is a gate the test settles by hand, so "the shell waited for X" is "no shell before X settled, shell after" — no clock, and ~1 s instead of ~2 s of real timers.

PII surface (RFC 08)

"Values in records — the PII surface": the complete list of record and finding fields carrying user data — ChangeRecord/HeldWrite previews (strings cut at 40 chars), ChangeOrigin/InteractionRef.target text (up to 30 chars of textContent), navigation to/from/params, the responsiveness findings' data.interaction/data.navigation, and data.error on the server error findings (the error as thrown, deliberately unsanitized — the wire got the generic message so the observer could see the real one; connection strings land here). And what is safe by construction: re-runs, and the runtimes' records, which keep args/results/thrown values on live. This is Solid's half of plan open question 6; the adapter's half is scrubbing per this list.

Verification

signals 2736 · solid 623 + 825 types · web 991 + 185 types · diagnostics 34 · universal 56. Size caps within limits. Changeset: @solidjs/signals, solid-js.

[changeset-bot]

🦋 Changeset detected

Latest commit: cc0396b

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
solid-js	Patch
test-integration	Patch
@solidjs/web	Patch
@solidjs/element	Patch
@solidjs/h	Patch
@solidjs/html	Patch
@solidjs/universal	Patch
@solidjs/babel-plugin	Patch
@solidjs/compiler	Patch
@solidjs/diagnostics	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 35122897082

Warning

No base build found for commit 59c4192 on next.
Coverage changes can't be calculated without a base build.
If a base build is processing, this comment will update automatically when it completes.

Coverage: 71.46%

Details

• Patch coverage: No coverable lines changed in this PR.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

Requires a base build to compare against. How to fix this →

---

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 not alter performance

✅ 175 untouched benchmarks

---

_{Comparing cross-package-fields (cc0396b) with next (59c4192)}