Give forked backend workers distinct socket-owner identities

[masenf]

When an app is imported before a production server forks, every worker inherits the same token-manager instance ID. A worker receiving a backend-initiated delta then mistakes another worker's live socket record for its own stale record and drops the update.

Assign a fresh socket-owner identity at event-processor startup, after the fork and before events are processed. The change adds no per-event PID checks or allocations. Addresses prerelease finding 003; the defect is verified on the forked-worker path and is pre-existing.

Validation:

• A new test forks the already-created app into two workers. Before the fix it fails because publishing the remote delta returns false; afterward both workers have distinct IDs and publish the expected payload to the owner's channel.
• A separate probe using two forked processes and a temporary real Redis server delivered all six backend-initiated updates through EventNamespace.emit_update, with different parent/owner/sender identities. All probe processes were stopped.
• App, token-manager, shared-state, and lifespan suites: 201 passed, 5 skipped. Ruff checks pass.

Reviewed the startup ordering, stable identity during event processing, and preservation of stale-local-socket rejection. This is independent of the production preload optimization in #7079, which makes the same lifecycle boundary particularly relevant.

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge, with startup ordering and cross-worker routing covered by a targeted regression test.

Summary

• Adds a focused fork-based regression test covering distinct worker identities and remote delta publication.
• Refactors token-manager identity generation into a reusable startup reset method.
• Adds a user-facing bugfix news fragment.

_{Reviews (1) · Last reviewed commit: "fix(tokens): assign socket ownership ide..."}

[cubic-dev-ai]

No issues found across 4 files

_{Re-trigger cubic}

[codspeed-hq]

Merging this PR will not alter performance

✅ 40 untouched benchmarks
⏩ 8 skipped benchmarks¹

---

_{Comparing codex/prerelease-003-worker-token-identity (312c863) with main (074a818)}

Footnotes

1. 8 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩