If the user's environment is stopped for some reason, ContainerShell finds that container and attempts to run an exec command against it. That's dumb because you can't start a process in a container that is not running. The smarter thing to do is to ensure the container is running in the first place. The logic is already there, it's just a small refactor to fix this bug:
https://github.com/willnx/container_shell/blob/master/container_shell/container_shell.py#L125