Replace-run hooks can't report a timeout outcome, so after-run hooks run after a provider-side --timeout #473

Found building labelle-ios v0.1.0 (labelle-ios#1, RFC #471 I1).

When a replace run provider enforces run.timeout_ms itself (e.g. labelle-ios stops the simulator app), it can only exit 0 or non-zero. The CLI's RunOutcome (provider_hooks.zig:463-518) isn't reachable from a hook, so a timed-out provider run looks like a clean exit and the after run hooks run. The CLI's own desktop watchdog skips them on timeout.

Options:

• a documented exit code (or a small outcome file in the context, e.g. run.outcome_file) through which the replacement reports timeout, so core applies the same after-run rule;
• or core enforces --timeout around the replacement itself.

The first is additive (contract 1.3 minor or 1.4).

https://claude.ai/code/session_01R2K5sJDL5Msn1uRTkybiDP