I can imagine a use case. For complicated queries, you end up performing a Cronenberg on your data in order to smash it all together with large join chains, or

I can imagine a use case. For complicated queries, you end up performing a Cronenberg on your data in order to smash it all together with large join chains, or even creating new json columns that effectively return other "tables". If its performance critical, this is often the best option. For most queries, a single table makes perfect sense. But in the real world things aren't usually so neat.