Thanks for the details — that error is expected: the worker tried to create a Kubernetes Job and the API rejected it because the memory request exceeded the limit. Prefect reports that as a Crashed flow run. About the UI “line progressing”: - In Prefect 3, the run-time bar only “ticks” while a run is in RUNNING. Once a run is in a terminal state (CRASHED/FAILED/COMPLETED/CANCELLED), the duration should freeze. - If you’re seeing it continue to progress, it’s usually one of: 1) A short lag before the state flips to CRASHED (the UI polls about every few seconds), 2) The page is still showing cached data and needs a hard refresh, 3) The backend is still reporting the run as RUNNING (e.g., state transition didn’t persist yet), 4) A frontend bug. Quick checks - Verify the run’s state and duration via CLI:

prefect flow-run inspect 480ef173-4c90-46ed-b187-bf8c663831fb

Confirm: -

state_type

is CRASHED (not RUNNING) -

estimated_run_time

is not increasing between runs of the command - If the CLI shows CRASHED and a fixed duration but the UI still animates after a hard refresh, that points to a UI bug. If you can share: - Prefect version for your worker and server/Cloud workspace - Whether you’re on Prefect Cloud or self-hosted - A screenshot (or just confirm if the run header shows CRASHED while the bar keeps animating) - The run ID (you’ve provided it) and approximate time you observed this …I can try to reproduce and escalate. I didn’t find an existing issue or discussion on this behavior; if it’s reproducible, we’ll get a bug filed. In the meantime, a hard page refresh or navigating away/back to the run typically stops the animation once the CRASHED state is reflected.