Activities that specify a valid timeout duration or datetime are likely implemented in a similar manner to wait and onAlarm activities with an expiration date for the underlying work item object. If the node that scheduled these activities with the scheduler goes down (either through graceful shutdown or abrupt termination), all these activities must be rescheduled with the scheduler upon server restart.

It is not possible to have a single node (the master node) in the cluster be responsible for rescheduling these activities upon node shutdown.