It makes only if you're doing a few operations and all of them are on tables. As soon as people need to mix up table operations and other code (custom functions

It makes only if you're doing a few operations and all of them are on tables. As soon as people need to mix up table operations and other code (custom functions and calculations, REST API calls, linear algebra etc) in the same script, it makes a lot more sense to use a dataframe and have all of your code in the same language. If there's a way in which sqlite is underrated by data scientists, it's as a storage format - it's a lot more reliable than passing around csv files.