I also do that for convenience. It helps a lot in many cases. In other cases I might have tables that may grow into the millions of rows (or hundreds of millions), then I'd absolutely not use UUID PK's for those particular tables. And I'd also shard them across schemas or multiple DBs.