PersistentActor enters infinite restart loop when RecoveryCompleted handler throws

[He-Pin]

Motivation

When a classic PersistentActor's RecoveryCompleted handler throws an exception, the actor enters an infinite restart loop instead of stopping. This is inconsistent with every other recovery failure path, which all explicitly call context.stop(self).

Current Behavior

In Eventsourced.scala:786-814, when RecoveryCompleted handler throws:

1. finally transitToProcessingState() runs — changes state
2. Exception propagates to outer catch which calls returnRecoveryPermit() and rethrows
3. No context.stop(self) is called
4. No onRecoveryFailure() callback is invoked
5. Default supervision (Restart with unlimited retries) causes infinite restart loop
6. Same journal events replayed → same exception → forever

Comparison with Other Recovery Failure Paths

Failure Scenario	onRecoveryFailure()	context.stop(self)	Result
ReplayedMessage throws	Yes	Yes	Actor stops
ReplayMessagesFailure	Yes	Yes	Actor stops
RecoveryTick timeout	Yes	Yes	Actor stops
RecoveryCompleted throws	No	No	Restart loop

Expected Behavior

The RecoveryCompleted failure path should be consistent with other recovery failure paths — call onRecoveryFailure() and context.stop(self) to prevent the infinite restart loop.

Reproduction

Confirmed by existing test RecoveryPermitterSpec.scala:174-197 which observes 5 consecutive restarts and must manually call system.stop() to break the loop.

Environment

• Pekko Persistence (any version)
• Eventsourced.scala:786-814

References

None