cursor agent hit a flaky e2e, flipped --headed, then timed out on page.waitForSelector. task died. chrome didn't.
activity monitor showed three Chromium Helper processes at ~1.1gb each. fans at 2am berlin. i had to pkill -f Chromium before the m2 would cool down.
anyone pinning headed to a debug profile only, or is everyone just hoping the agent cleans up?
left an agent overnight on a staging branch with prisma migrate dev . woke up to prisma migrations showing one row as finished at: null and the next migration refused to start with P3009. the sql par…
keyed actions/cache on ollama models ${{ hashFiles('Modelfile') }} . Modelfile only had the model name, not a digest. runner came up with a half pulled qwen2.5:7b from last tuesday. fixture hit Conne…
left a Roo Code session running Friday night against a LangSmith eval suite. one case was flaky — TimeoutError: run exceeded 60s — and the agent just… kept retrying it. no max attempts. no spend cap.…
2 comments
Join the discussion
Log in to comment.
yeah we banned headed in the agent prompt after it left 6 chrome processes overnight and the runner disk filled with crash dumps.
debug profile only, and a post-job
pkill -f chromiumin gha. ugly but cheaper than another $0 mac mini melt.same class of bug as mlx leaving orphan python workers after an OOM. agents never own process trees.
i run headed only on a spare win box now. m2 air stays headless or it starts thermal-throttling mid-refactor.