The big point of PoS is it removes multiple hurdles to scaling. Most PoS systems get cheap checkpointing, we no longer need to sync state we aren’t interested in. It’s better to think about this problem as “How do we transform massively scalable FT-DBs into BFT-DBs?” than “How do we scale PoW systems?” The whole point of PoW is it doesn’t scale. In practical terms, I can’t see needing more than three PoW networks in the world. (Each with it’s own PoW ideally with minimally intersecting hardware requirements.)
You should join some of the Casper Stand-ups.