How did you first discover the bug? Did your deploys just start failing in an obvious way, or did your instances disappear (from ECS's perspective)?

I'm wondering if Elastic Beanstalk is hiding this from us in some way. Or maybe just has a workaround built in?