[bushbaba]

For smaller datasets (100s of thousands of rows) I don’t see why you wouldn’t just use json columns with generated column/index where needed

[bilekas]

Regardless of the number of rows, it doesn't really matter, there are useful cases for where you might be consuming json directly, so instead of parsing it out into a schema for your database, why not just keep it raw and utilize the tools of the database.

It's a feature, not a replacement.