fix(desktop): sync $activeSessionId atom after session-recovery auto-retry

[luxles]

Problem

When prompt.submit returns "session not found" (stale runtime session ID after Gateway restart, WS orphan reap, or context compression), the frontend auto-recovers by calling session.resume to get a fresh runtime ID, then retries the submit.

However, the recovery only updates activeSessionIdRef.current (a ref) without calling setActiveSessionId() (the React atom).

The next submitPrompt call gets sessionId from useStore($activeSessionId) — the atom, which still holds the old stale runtime ID. So every other message fails with "session not found", creating a retry loop.

Fix

Add setActiveSessionId(recoveredId) on the recovery path alongside the ref update.

File: apps/desktop/src/app/session/hooks/use-prompt-actions.ts

Related Issues

Fixes #51058
Related: #38704, #44183, #47709

[teknium1]

Thanks for identifying the atom/ref divergence; the premise remains valid on current main.

Problems

• The submitted diff changes apps/desktop/src/app/session/hooks/use-prompt-actions.ts, but the live retry branch was extracted by 7337248a4c182d65b6c101acac1e21be5101c4cc and now lives at apps/desktop/src/app/session/hooks/use-prompt-actions/submit.ts:298-309. The current branch still only writes activeSessionIdRef.current at submit.ts:306, so the fix must be transplanted there.
• The same ref-only recovery pattern also appears in submit.ts:226-233 and index.ts:548-557; please keep the atom and ref synchronized across these current recovery paths.
• index.test.tsx:1050-1095 verifies the retry RPC sequence but not $activeSessionId; add a regression assertion for the recovered atom value and a subsequent submit.

Suggested changes

• Import and call setActiveSessionId(recoveredId) beside the current ref updates, then extend the existing recovery test.

Automated hermes-sweeper review.

[teknium1]

This is the correct state transition, but current main extracted this retry branch to apps/desktop/src/app/session/hooks/use-prompt-actions/submit.ts:298-309 (commit 7337248a4c182d65b6c101acac1e21be5101c4cc). Please transplant the atom update there; this file is no longer the live implementation.

[luxles]

Updated per review feedback

Rebased on latest main and transplanted the fix to the refactored file structure:

Fixed 2 recovery paths:

• submit.ts:368 — setActiveSessionId(recoveredId) added before the retry submit
• index.ts:559 — setActiveSessionId(recoveredId) added before the interrupt retry

Added regression tests:

• index.test.tsx — asserts $activeSessionId.get() === RECOVERED_SESSION_ID after both prompt.submit and session.interrupt recovery paths

cc @teknium1

[OutThisLife]

Superseded by #81261.

Thanks for this — the $activeSessionId desync after recovery is carried over there, with you credited via Co-authored-by.

Consolidating because this is one bug class across several PRs: main had three hand-rolled copies of the same stale-session recovery (prompt.submit, session.interrupt, session.redirect), and the RPCs that never got a copy — attach, /compress, checkpoint restore, tile actions — each failed the same way. #81261 makes it one resolver and routes every path through it, so the next session-scoped RPC we add inherits the fix instead of needing a fourth copy.