Sharded message workers

[PerWiklander]

I want messages to the same entity to be handled by the same worker, to allow holding state in memory in the worker, efter having rehydrated the state.

Think Akka cluster sharding and event sourced actors.

I have seen this and it looks like a good way of dividing a stream of messages into partitions:
https://docs.nats.io/nats-concepts/subject_mapping#deterministic-subject-token-partitioning.

But does NATS (or a pattern that I can apply on NATS) help me in any way with running the workers or rebalancing the workload if for example one of three worker nodes goes down?