ReplicaSetAwareService implementations have the ability to run in drain mode and some of the ones that we have take advantage of that.

We should add some logic in the ReplicaSetAwareServiceRegistry which keeps a tally of how much time each service spends in drain mode and then logs a warning if:

1. The total time all services take is more than some threshold (say 500ms)
2. Either one of the services takes more than some threshold (say 250ms)

If there are other events on the RSAS that have the potential of causing unavailability if they take too much time, we should add warnings for them as well.