fix: pooled event processor - coordination tasks not scheduled when releaseClaim throws exception

[fabio-couto]

Basic information

• Axon Framework version: 4.5.x
• JDK version: 11.0.13

Steps to reproduce

When using a RDBMS as Event Store, Event Bus and Token Store, and the database goes offline, the scenario below occurs:

1. The CoordinationTask from Pooled Event Processor cannot read more events from stream, so it starts the abortAndScheduleRetry logic;
2. In abortAndScheduleRetry, the application tries to release claim in token store, but as the database is offline, this throws another exception;
3. The completable future chain are not handling exceptional completation, and because it, the task reschedule and the wait back off logic are not being executed;
4. When the database comes online again, as the reschedule is not happening, the application does not recover by itself, requiring a restart of the JVM.

Expected behaviour

The coordination task reschedule should execute always (respecting the error wait back off intervals).
So, when the DB comes online again, the application can reconnect and continue processing events.

Actual behaviour

The application does not recover by itself, does not throw exceptions, and does not generate a log record reporting this situation. Also, the related event processors don't handle the new events, requiring a restart of the service.

[CLAassistant]

All committers have signed the CLA.