Does according to author, NATS is less reliable, if you want guaranteed ordering and to have multiple instances of worker, because NATS partitions lack "auto-balance" like in Kafka?<p>I would assume that NATS can be deployed in some Kubernetes cluster, so it
would not be uncommon to have 3 or 5 workers. In this case, what would happen, if someone wanted guaranteed ordering and used that deterministic subject token partitioning?<p>Let's assume that some worker would crash due to lack of storage space and it would keep restarting, so what happens to messages on this partition? Can they be processed, if specific worker is not available? Is it possible to react to this event and manually reassign these partitions to other workers? In case that there's no event, then maybe it's possible to write some script and run CronJob to manually check, if rebalance is needed?