Clear fatal-error interrupt window unconditionally on body exit

[masenf]

Type of change

• Bug fix (non-breaking change which fixes an issue)

Changes To Core Features:

• Have you added an explanation of what your changes do and why you'd like us to include them?
• Have you written new tests for your core changes, as applicable?
• Have you successfully ran tests with your changes locally?

Description

Follow-up to #6990 (merged while this review round was in flight), addressing the two review threads that were still open there:

1. cubic's finding is valid: run_concurrently_context cleared in_body only on the normal post-body path and in the KeyboardInterrupt handler. If the with-body raises any other exception, in_body stays True — and since the executor shuts down with wait=False, a task failing after the context has unwound could still deliver a stray SIGINT into unrelated caller code. The fix wraps the yield in a try/finally that clears in_body however the body exits, replacing the two scattered clears — a net simplification of the mechanism, not an addition (per the review discussion, no extra guarded setup path).
2. greptile's nit: the hang regression test's bare 5-second assertion threshold now carries a comment explaining its relationship to the body's 10s wait.

Adds test_run_concurrently_context_no_interrupt_after_body_exception, which pins the escaped-interrupt case (late-failing task + body exception → the body's own ValueError propagates, no stray KeyboardInterrupt).

Test Plan

• uv run pytest tests/units/utils/test_processes.py -k run_concurrently: 4 passed (the new test plus the three from Fix reflex run hanging after fatal error in the frontend thread #6990).
• uv run ruff check / ruff format --check clean on changed files; uv run pyright 0 errors on changed files.
• reflex-release changelog-check --base-ref origin/main passes.
• The two test_is_process_on_port_* failures in this sandbox are pre-existing on clean main here (container network stack), unrelated to the diff.

🤖 Generated with Claude Code

https://claude.ai/code/session_01EMjBXPozsNeQNSBZecNH8x

---

Generated by Claude Code

[masenf]

Note: this was opened moments before your replies landed on the #6990 threads ("if we're going down, we're going down" / "test code"), so calling out the delta explicitly: executor.shutdown(wait=False) means a worker can still be running after the context unwinds, so when the body exits via a non-KeyboardInterrupt exception (where in_body stayed True), a late task failure could deliver SIGINT into caller code beyond the context — the new test pins that concrete case, and the diff is a net simplification (one finally replaces the two scattered clears). If you still consider the edge not worth carrying, feel free to close this — no strong attachment.

---

Generated by Claude Code

[greptile-apps]

Greptile Summary

The PR clears the fatal-error interrupt window in a finally block whenever context entry or the with-body exits.

• Consolidates in_body cleanup across normal, exceptional, and pre-body failure paths.
• Adds regression coverage for late worker failures after body and pre-body exceptions.
• Documents the timing thresholds used by the existing interruption test.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
reflex/utils/processes.py	Moves interrupt-window cleanup into an unconditional nested finally, covering all body-exit paths.
tests/units/utils/test_processes.py	Adds coverage for late worker failures and documents the existing timeout relationship.
news/+interrupt-window-followup.bugfix.md	Records the corrected late-interrupt behavior in the changelog.

_{Reviews (3): Last reviewed commit: "Clear the interrupt window when the pre-..." | Re-trigger Greptile}

[codspeed-hq]

Merging this PR will not alter performance

✅ 32 untouched benchmarks
⏩ 8 skipped benchmarks¹

---

_{Comparing claude/rel-fix-interrupt-window-followup (276300f) with main (6c6a398)}

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. ↩

[cubic-dev-ai]

All reported issues were addressed

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>