That warning usually means the worker is shutting down and Prefect is waiting for its background event-sender to finish flushing queued events to Prefect Cloud. So in most cases, it is not harmful by itself. What

EventsWorker

is doing - Prefect workers emit events like worker started/stopped, flow run state changes, etc. - Those events are sent asynchronously by a background component called

EventsWorker

- When the worker is stopping, Prefect tries to drain that queue before exiting - If there are still unsent items, you’ll see:

WARNING | EventsWorker - Still processing items: 1 items remaining...

Why you might see it “always” on Windows A few likely reasons: - Normal shutdown behavior: if you start/stop the worker often, there’s almost always at least one final event left to send - Slow network/API connectivity to Prefect Cloud, so the final event takes a bit longer - Windows thread / I/O timing differences, which can make this more visible than on Linux - Older Prefect version: newer 3.x releases include fixes around event worker shutdown behavior Is it a problem? Usually no if: - the worker still starts and runs jobs normally - you only see this during shutdown/restart - it eventually exits cleanly It’s more concerning if: - it repeats continuously while the worker is running - shutdown hangs for a long time - you also see connection/timeouts to Prefect Cloud What I’d try 1. Upgrade Prefect to the latest 3.x:

bash
pip install -U prefect

2. Check whether it only appears on shutdown - If yes, this is likely just a noisy but benign warning 3. Enable debug logs to confirm what event is stuck:

bash
set PREFECT_LOGGING_LEVEL=DEBUG
prefect worker start --pool YOUR_POOL --type process

4. Check connectivity to Prefect Cloud - Make sure the machine can reliably reach your API URL - Confirm

PREFECT_API_URL

and

PREFECT_API_KEY

are set correctly 5. Optionally reduce event backlog / tune queueing You can try a bounded queue:

bash
set PREFECT_EVENTS_WORKER_MAX_QUEUE_SIZE=2000

If this is specifically happening when the worker launches rather than when it exits, that’s less typical and I’d want to narrow it down. A few quick questions: - What Prefect version are you running? - Do you see this once at startup, or does it repeat every few seconds? - Are you running

prefect worker start

in a normal console, as a Windows service, or under some supervisor? - Do you also see any timeout / connection warnings nearby in the logs? Helpful docs: - Workers documentation - Work pools and workers - Settings reference If you want, paste the surrounding startup logs and I can help pinpoint whether this is just shutdown noise or an actual connectivity issue.