Also what I think. And it feels like partitioning would imply much more work and potential future implications rather than figuring out what was the issue with some queries or indexes. Postgres should be able to handle billions of rows.

And if the issue was with the queries per second, it would feel to me like it would still be less disruptive to just have some sort of replication going on for reads.

And partitioning really makes sense to me only if you are going to have it on multiple boxes, because otherwise I don't think there really should be many performance gains compared to just indexing?

I'm not seeing if/how they put it on multiple boxes?