The wait interval is how long the broker will suspend allocations while waiting for all the engines to failback to the primary. Once the interval expires any stacks/components in the currently active policy that should be running but the broker thinks are not will be attempted to be activated. If those components are actually running on engines that have not yet failed back they will be deactivated when they do login and there will be a chain reaction of deactivating & activating & re-activating components for the next few hours until everything stabilizes. If a user wants to avoid this, they should set the "Failover Wait Interval" to something like 1 hour. The downside of a 1 hour wait is that the allocator is suspended for that period.

The default is 120 (2 minutes) which is fine for a small Dev/Test environment but needs to be much longer for a reasonable number of engines.