Calls an RPC with retry logic and exponential backoff.

Accepts partial failures and uses adaptive timeouts to handle network delays and temporary unavailability.

Evaluates success rate and determines if operation succeeded.

Accepts partial success for eventual consistency:

• = 80% success: Operation considered successful

• = 50% success: Degraded but acceptable

• < 50% success: Critical failure

The system relies on self-healing to converge to correct state.

Gracefully stops an actor with timeout and monitoring.

Gives the actor time to clean up before forcing termination. Tags the shutdown reason with the rebalancing epoch for debugging.

Stops actors in parallel with timeout.

Uses Task-based parallelism to stop multiple actors efficiently, with a global timeout to prevent indefinite blocking.