Used watches for better resiliency upon failure #2
Description
Leverage watches in etcd to account for failures of the controller, docker, machine and network on restart.
If a machine in the leadership ring goes down, this should be updated within 1 second so that any machine subsequently coming up replaces it in the leadership ring.