fix(agents): keep REPL alive on telemetry stalls

[FluffyAIcode]

Summary

• apply a short timeout to optional REPL telemetry calls
• warn and continue inference when benchmark or prefill metrics temporarily fail
• retain the latest stats snapshot and relax KV metric gates only while telemetry is degraded
• compute the final Generator/Critic summary locally if dashboard persistence fails
• preserve remote benchmark persistence whenever the API remains available

Root cause

The REPL synchronously treated benchmark creation as mandatory. A transient dashboard API stall therefore raised TimeoutError before inference began and escaped the per-turn error boundary. The dashboard has since recovered and currently responds in 1–3 ms, but telemetry must remain non-critical.

Tests

• fail-open timeout regression and warning behavior
• Agent GAN REPL/demo/report suite — 15 passed
• full platform-neutral suite — 858 passed, 1 skipped
• Python compile and diff checks pass

Made with Cursor

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.