Zanie Blue

I suspect what is happening here is that Dask is executing the task run multiple times and we are aborting the second run then attempting to retrieve the result from...

> Are there any timeout / magic config numbers related to anyio or prefect.engine at the task level that might be set to 330 seconds? Not that I know of.

@space-age-pete note this gets at the nuance we are concerned about. We do not support cancellation of running flows at this time. If the flow has crashed and was not...

Have you looked into https://docs.aws.amazon.com/AmazonECR/latest/userguide/registry_auth.html / https://github.com/awslabs/amazon-ecr-credential-helper/issues/207 ?

Are you able to pull the image with `docker pull `? We need to support private image repositories. I'm trying to narrow down where the authentication is not being included....

So... this is actually a `docker-py` issue: https://github.com/docker/docker-py/issues/2256 It looks like we might be able to incorporate a workaround though.

I believe this will require us to add auth to the `DockerContainer` object, probably through a `DockerRegistry` block.

Thanks for the follow-up! I'll investigate that thread issue, I knew something weird was lurking in there as I've seen some errors in the past.

@a6502 it's going to be challenging to ensure that the event belongs to the correct loop here. Can you share a minimal example that is improved by this implementation so...

If you could try changing the concurrent task runner implementation to just call `time.sleep(0)` after a `anyio.sleep(0)` and let me know if that improves performance in your workflow as well,...