It's not gonna need that for a long time with the anti-bot anti-spam anti-troll measures in the reputation coefficient system. For future, it's next iteration is already planned, to be based on SporeDB: https://github.com/SporeDB/sporedb which is not just multithreaded but has multiple parallel backends, already, the initial one being RocksDB, which is used by facebook. I think Facebook knows about scalability...