Thanks for flagging this — that shouldn’t happen. In Prefect 3.x,

prefect deployment run --watch

streams events/logs and exits as soon as it sees a terminal state (Completed/Failed/Cancelled/Crashed), then waits ~3s for any “straggler” logs before closing. If your terminal stays stuck after the flow succeeds, the watcher likely didn’t receive the terminal event (e.g., a transient event stream issue, proxy buffering, or self-hosted background services not fully running). A few quick things to try: - Add a timeout so the watch cannot hang indefinitely

prefect deployment run 'FLOW/DEPLOYMENT' --watch --watch-timeout 300

This will exit after 5 minutes even if a terminal event isn’t received. - Confirm you’re on a recent 3.x

prefect --version

If you’re on an older release, upgrade:

pip install -U prefect

There have been steady improvements to the event/log subscriber. - If you’re self-hosting Prefect Server, make sure Background Services are running and healthy (they’re responsible for event delivery). If you’re on Prefect Cloud, check Prefect Status if this happens intermittently. - As a workaround if you need a non-blocking run: - Run without

--watch

, then inspect or fetch logs by ID:

# Create the run (no watch)
    prefect deployment run 'FLOW/DEPLOYMENT'

    # Inspect or fetch logs once you have the run ID
    prefect flow-run inspect RUN_ID
    prefect flow-run logs RUN_ID --tail

If you can share a bit more detail, I can help narrow it down: - Output of

prefect --version

- Prefect Cloud or self-hosted Server? - The exact command you ran - Does it happen every time or intermittently? - If self-hosted, logs from the API/Background Services when the watch hangs - Whether you’re behind a proxy or using a terminal that may buffer long-lived HTTP streams If it’s repeatable and you’re on a recent 3.x, we can dig deeper (and I can open an issue with the details).

Thanks for the details — given you’re on Prefect Cloud and 3.6.8 and it happens every time, this does look like a regression in the watch event handling. Quick mitigations you can use right now: - Add a timeout so the watch can’t hang indefinitely

prefect deployment run 'FLOW/DEPLOYMENT' --watch --watch-timeout 300

- Or run without

--watch

and then inspect/logs by run ID:

prefect deployment run 'FLOW/DEPLOYMENT'
  # Copy the run ID from the output or UI
  prefect flow-run inspect RUN_ID
  prefect flow-run logs RUN_ID --tail

To help us diagnose and file a bug, could you share: - One or two example flow run IDs where this occurred - Approximate timestamp/timezone - Your OS and shell (e.g., macOS zsh, Windows PowerShell) - Whether you’re behind a proxy/VPN - What you see at the end of the watch output (e.g., does it show “Completed” and then hang, or does it just stop streaming?) If you can re-run once with debug logs and paste the last ~50 lines, that’s super helpful:

export PREFECT_LOGGING_LEVEL=DEBUG
prefect deployment run 'FLOW/DEPLOYMENT' --watch --watch-timeout 300

What I’ve verified so far: - The CLI does support

--watch

and

--watch-timeout

(event-stream based). - I don’t see an existing public issue for 3.6.8 hangs with

--watch

. - This behavior is consistent with the watcher not receiving the terminal state event from the stream (Cloud) and waiting forever unless a timeout is set. If you’d like, I can open a GitHub issue on your behalf once you share a reproducible run ID and environment details. Otherwise, feel free to file one directly here: Open a Prefect issue. I’ll keep an eye on it and route it to the right folks. Also, if this suddenly starts/started happening today, you can check Prefect Status just in case — but since it’s every time for you on 3.6.8, I’m leaning toward a client-side regression.