Atlantis

It should wait to drain connections before it kills the container, I think.

you need to look at memory utilization and see if you are hitting any limits

some big plan can take up a lot of memory

usually I set the fargate task to at least 1/2+ of that I need

Memory seems fine.  This would be for an `apply` that takes 20 minutes while it waits for a MQ broker to reboot (or EKS cluster creation, etc) and Fargate launches a new task because the container is updated (maybe from an `atlantis.yaml` update, or a new version).  Sounds like Fargate will wait until the container exits though so it might not be a problem.  I can do some testing to validate.