The high availability (HA) manager always calls the partitionLoadEvent before the partitionUnloadEvent, and the invocations occur serially for the same partition (they are never called concurrently for the same partition). The HA manager maintains a dedicated thread pool for invoking these methods. When N partitions are activated, the load events are delivered to the application using this thread pool; therefore, if the thread pool has 10 threads, then at most 10 partitions are activated/deactivated at once. If there are 100 partitions to activate, then the 100 activation method calls are queued to the thread pool and the thread pool delivers the events to the application 10 at a time. This is approximate and depends on CPU count, and the operating system.