You can set the following option to enable in-memory event handling:

This default value is true. When true, the Task Engine does not use the JMS queue for processing task events and all events are processed locally on the same JVM instance where they were originated. This produces less database load and thus speeds up processing. The disadvantage of using this setting is that there is a bigger chance for events to be left unprocessed if the system shuts down unexpectedly or fails.

To overcome this, you can use a related option to specify if task events are to be maintained in system memory, or persistently stored to disk:

The default value is false (use system memory).

When persisting task events to disk (true), the processing of each task change event to disk imposes a performance penalty, but adds quality of service. For example, if a system shuts down unexpectedly or fails while there are unprocessed events in the queue, the system can recover them from the disk store and continue processing when restarted.

When task events are stored in memory (false), these events will be lost upon shutdown or failure. This can result in queued tasks not getting assigned (that is, the tasks were queued but not yet assigned when the shutdown or failure occurred).

The -Dtask.event.lightweight.persistent option is active only when -Dtask.event.lightweight is set to true.