PR #1109 only handles errors inside stream_input - 3+ other root causes of hanging query() not addressed

[knoal]

Problem

PR #1109 (fix: surface prompt stream errors instead of hanging query()) only handles errors inside stream_input's own try/except. The "hanging query()" symptom has at least 3 other plausible root causes not addressed.

Impact

• The fix may not address the actual hang root cause in many cases
• A real hang could still occur in the unhandled scenarios
• The test relies on anyio.fail_after(5) to convert a hang into a pass - brittle and not really a failing-first test

What was audited

MCE audit Tcas1109 (2026-07-19) caught 3 HIGH attacks via F-C:

1. The test does not actually exercise the original bug on a failing-first basis - before the fix, stream_input() swallowed the exception and returned; the CLI would then wait for input
2. P1's confidence of 0.85 is overstated - the fix only handles errors inside stream_input's own try/except
3. P4 confidence (0.6) is too high and the test is brittle - it relies on anyio.fail_after(5) to convert a hang into a pass

Recommended fix

1. The author should:
   a. Identify and address all 3+ root causes of the hanging query() symptom
   b. Write failing-first tests that exercise each root cause
   c. Avoid relying on anyio.fail_after to convert hangs into passes

Cross-references

• MCE audit: Tcas1109 (2026-07-19), 0/4 HIT, Brier 0.55
• F-C analysis: /home/alex/.hermes/mce/fc-critiques/Tcas1109-with-diff.json
• PR: fix: surface prompt stream errors instead of hanging query() forever #1109