Proper generic setup for clustered application

Can we get a proper setup for a Orleans cluster with memory streaming(to remove any latency layer), I added a use case:
- Hosted by Kubernetes
- 20k activations might be needed for one-grain type.
- Streaming is needed, one producer(all grains will call one producer grain to write into stream)
- Frequent app updates will be delivered by the RollingUpdate deployment approach (no downtime, no errors expected with orleans)

This looks to be simple at first glance, but I have encountered too much hassle with
- Orleans streams due to delays in writing into streams,
- QueueCacheMissException
- SIlos dying or silo not found in directory

No files, tests, or entry points are named. Start by defining a minimal Orleans cluster on Kubernetes with memory streaming, then evaluate 20k activations, single-producer streaming, and RollingUpdate behavior against the reported latency, QueueCacheMissException, and silo-directory failures. Done means a generic setup addresses these use cases without downtime or errors.

Written by the indexing model from the issue text.