Usage limit killed scheduled agent jobs for 19 hours — team shares 4 fixes so agents monitor themselves
lilythemoon54 · reddit · 2026-09-21
A team's scheduled agent jobs all share one account; when it hit the weekly usage limit in mid-September, every model-booting job failed to start for about 19 hours while plain-Python jobs on the same schedule kept running — the agent that would have reported the failure never booted, so nobody noticed.
Their post-incident changes:
- Every scheduled job starts with a plain-code gate that decides whether the model is needed, so quiet days cost no quota;
- The gate matches the exact limit message and labels it as quota, not a bug, avoiding generic retries slamming the same wall;
- Expired logins get their own check since the fix requires human sign-in;
- Routine jobs default to a cheaper model to stretch the weekly allowance.
The core lesson: "whatever tells you an agent failed can't be the agent" — the heartbeat must live outside the agent.
More from coding & agent
- HarnessRouter open-sources one API to run Codex, Claude Code and 14 agent harnesses — gaganghotra_ · 2026-09-21
- ECDYSIS paper: patch agent harnesses by failure patterns, not failure counts — rohanpaul_ai · 2026-09-21
- EU LLM apps: developer maps the 5 blockers between prototype and paid launch — felix_baron · 2026-09-21
- Coding agent UX gripe: Astra says "PR is up" without linking to it — altryne · 2026-09-21
- God's Eye View: open-source spy-satellite simulator with real data hits 39.5k GitHub stars — alex_verem · 2026-09-21
- GEPA Prompt Optimization Lifts Jev's F1 From 69.1% to 79.7% on Medical Literature Task — matei_zaharia · 2026-09-21