2.0.0-rc.8 optimistic update keeps waiting after its last async reader unmounts

[GabbeV]

Describe the bug

An optimistic update keeps waiting for an async memo after its last reader is unmounted. The equivalent normal signal update releases immediately.

Your Example Website or App

https://s.olid.uk/id/ivHMtOauSn2E1c4bfvpFDg

Steps to Reproduce the Bug or Issue

1. Wait for initialization.
2. Click Run.
3. Details disappears after 500ms, but Value stays 0.
4. Value becomes 1 when the unused request finishes. The action is still pending.

Expected behavior

I would expect the optimistic update to release when nothing observes the pending work anymore, like a normal signal update.

Screenshots or Videos

No response

Platform

• OS: Windows
• Browser: Firefox

Additional context

Possibly related to #3372, fixed by #3392, which wakes a parked transaction when its pending reader is disposed.

This case eventually completes; it does not lose the update as #3372 did. Is keeping the optimistic lane held after disposal intentional, or should it release here too?

Reproduced on next at 1643d2a and #3425 at 9fa294f.

[ryansolid]

Confirmed and fixed in #3434 — thanks for the reproduction.

Root cause: the lane held Value: 1 on the Details reader that observed details pending. A live action parks its transaction without a settle verdict, and the verdict was the only place dead reporters were pruned — so when the reader unmounted mid-action its registration stayed, and the lane kept waiting for a flight nobody observed anymore.

Rule now: the hold check prunes as it reads. The live-reporter test (sourceObserved: disposed, behind a fallback, or no longer reading the node → not live) is shared by the settle verdict, the lane hold and the landing. Your frames become 2500: Details: <unmounted> | Value: 1 at the unmount and Value: 0 when the action ends — the same as a plain signal write.

— Claude via Cursor

[ryansolid]

Fixed in 25c5064 (#3434), merged to next — ships in the next rc.

The lane held Value: 1 on the Details reader that had observed details pending. A live action parks its transaction without a settle verdict, so that reader's registration was never pruned when it unmounted, and the lane kept waiting for a flight nobody observed. The hold check now prunes as it reads: sourceObserved is the one live-reporter test shared by the settle verdict, the lane hold and the landing. The frame reveals at the unmount and reverts when the action ends — same as the plain-signal version you compared against. Pinned in optimistic-lane-release.test.ts.

— Claude via Cursor