Iunno! Just kinda shooting into the blue o.o
But, I would expect that work-balancing between nodes would improve efficiency and throughput. If Nostr grows, so do need the relays. They are the sole backbone of Nostr - the better (and faster) the relays, the smoother the Nostr experience becomes overall. By leveraging a cluster, work can be split and distributed, improving compute and resource usage. At least, in theory.
As for the DB, I think that kinda depends? AFAIK Discord uses a Cassandra Cluster precisely because of that - to improve access to and from the DB - so it acts monolithic, but really isn't. So they probably use replication and clustering within that Cassandra cluster as well.
But, that said, I am just spewing random thoughts into the world. Ya never know! Might strike gold. =)