In a highly available configuration, when the primary instance goes down, the back-up instance becomes the primary instance and processes the events and rules without any impact on performance. If the previous primary instance recovers from a failure, it becomes the back-up instance.