I'd be interested to know what proportion of users of dataframes are working at different orders of magnitude.

Most of my life I've had databases of like 1000. Now I have a big one of about 500K! So for me, speed is almost a non-issue. But that is my specific field.