Each Always-On Monitoring instance that receives an alert stores the alert and records an entry in one of 16 event notification queues. Ownership of these notifications queues is distributed among the running Always-On Monitoring instances.

A failover system notices when Always-On Monitoring instances are not up (and heartbeating) and switches ownership away from the instance that is down to other Always-On Monitoring instances that are up. The failover system then sends a JMS AQ message notifying all up instances of the change.