Comment 2 for bug 1677425

This is still failing sometimes in 2.3.

I think the issue is that the host machine is taking a long time to provision, and the container sits there in the "dying" state during that time. When the host finally comes up, its machine agent sees the container is dying and proceeds to remove it from state.

We should short-circuit the removal of containers when the host machine hasn't yet been provisioned.