We've encountered a problem twice with MongoDB that the host process is defunct/zombie, the container is still running, the pod is terminating permanently and the host process can't be killed with -9.

This is going to happen and we need a strategy to handle it going forward – maybe reboot the node is the only way?