we're working on a guide, but for now, you can refer to this blog post if deploying on Kub: https://authzed.com/blog/consistent-hash-load-balancing-grpc/
> Secundo, does that mean that every SpiceDB instance in the deployment is able to serve traffic and to serve dispatch requests from other instances? So that SpiceDB is highly-available, traffic being load-balanced between all the instances
Yes, that's correct.
> Tertio, can the datastore scale horizontally too? Eg can I have a Postgres read-replica and have SpiceDB make use of it for check and expand requests?
Currently, no. The recommendation if you want a truly horizontally scalable datastore is to use CockroachDB or the (beta) Spanner driver. Postgres driver only works with a single master, as that's required to maintain consistency