Hello!
I have a worker that is producing confusing CPU-time reports and statistics. When the CPU-time graph on the worker dashboard is viewed with hourly averaging (24h or greater timescales), it reports entire hours as averaging over 2000 ms.
If I look into those same periods at smaller timescales, I can find no single request approaching anything like those times:
My actual timeouts are pretty minimal, less than 0.02% of requests. In the 6PM spike in the 24h plot, we had thousands of requests processed by this worker. I would have expected a large number of timeouts and mayhem on the front-end if the 50th percentile was really at 2000 ms.