Custom actor partitioning (hash) scheme
As a SaaS developer I want to use actor partitioning to implement multitenancy, where all actors for a given client tenant are collocated within a single partition. (My understanding is that this will keep them all on a single node, while keeping replicas of that partition spread across other nodes in the cluster.)
One approach: make ActorId subclassable with a virtual implementation of 'GetHashCode'. Perhaps a better option would be a virtual implementation of 'GetPartitionKey' that by default calls GetHashCode. Either way, I could then create my ActorIds following a pattern (supported by my subclass's constructor) of "partition-key" and "instance-id".
Or even better, perhaps you could just support this scenario out of the box with a standard 'PartitionedActorId' class.