Presumably sharding is a lot easier than trying to debug lockups in individual postgres thread? It's well known, we've been doing it for at least 30+ years as an industry.