🎯 Changes

Closes #11903.

onCacheEvent bumped version for every cache event on the hook's hash. After a key switch, observer.setOptions in the options render effect emits observer events (observerAdded, observerOptionsUpdated, observerResultsUpdated), so the data projection ran a second time in a later flush. That second run is what Solid's diagnostics report as EFFECT_RELAY_TEAR.

version now moves only on state events (added, removed, updated). observerResultsUpdated still has to reach isStale, since it is how a query going stale by time is reported (two existing useQuery.test.tsx tests fail without it), so it bumps a separate signal that only isStale reads.

The new tests in useQuery-semantics.test.tsx pin one data-projection recompute per switch to a cached key (and back), counted through select: [2, 2, 2] before (observer + projection twice), [2, 2] after.

Checked locally with @solidjs/diagnostics@2.0.0-rc.13 (not added to the repo): the cached and switch-back scenarios from the issue no longer log EFFECT_RELAY_TEAR, with identical rendered output. Not covered:

• a memo reading query.isStale still recomputes twice on a key switch;
• the stale-key back-and-forth case from the issue comment (staleTime: 0) still logs ABANDONED_FLIGHTS and EFFECT_RELAY_TEAR (same output before and after this change).

Note: this PR description was drafted with AI assistance. The AI policy box is left unchecked for that reason.

✅ Checklist

• I have followed the steps in the Contributing guide.
• I have tested code changes locally with pnpm run test:pr, or these tests do not apply to this pull request.
• I have followed the AI contribution policy and fully understand the code in this pull request, including any code generated with AI assistance.

🚀 Release Impact

• This change affects published code, and I have generated a changeset.
• This change is docs/CI/dev-only (no release).