A lot of data scientists don't know SQL and don't understand why people use it. That said, there are cases where in memory workloads and certain manipulations are less encumbered by dataframes APIs. But in a lot of cases... They get abused by people who really do need to push themselves a little bit to learn something new.

You use both. Once your data fits comfortably in memory it's naive to try to build histograms, pivots and charts using pure SQL.

You're saying all Pandas usage (an incredibly popular library) is because people don't want to use SQL?

It's a broken argument with some truth too it. IE you can run a SQL query put the result in a dataframe to dump it to an interchange format. But in the same breathe... If you learn to use SQL a great deal of workloads often used via dataframes APIs kind of disappear, and in doing so learning a new language to use a new dataframes API isn't really worth it.

As far as I am aware, plenty of use cases can also be done via OLAP.