Restarts are not handled elegantly

[Others]

We need to consider what happens when the deployment manager or the worker nodes go down -- with no push, apps may be down permanently