One of the problems with spreadsheet usage stats is that a significant fraction of spreadsheets in the wild exercise nothing more than simple arithmetic; say monthly sums and a grand total of the yearly. One would be hard pressed to improve the usability of the system, when the user is doing so little data manipulation. Spreadsheets give the user 100% visibility of the data, but hide all the code. As long as you have negligible amounts of code this tradeoff works great. But i think most people would agree that spreadsheets are quite dangerous; so easy to remove a row without any trace of that mistake, and it is easy to make errors in formula ranges. The core functionality is really mostly unchanged since Frankston & Bricklin invented the category so long ago. I do think that an easier spreadsheet could thrive that has more safety checks against bonehead mistakes. So much of Excel's evolution has been to add rarely used complex features. I see Excel as the poor-man's database; it works remarkably well up to a million rows then it poops out. Having made a spreadsheet back in the day, it actually a pretty easy thing to program in the core, but all the graphical niceties make it an extremely complex beast to program.

I've been hearing from users that a million rows are only tractable if you're not doing any computation. Several people have told me that their spreadsheets crap out at 1k rows (!)

My coworker was just using Excel as a data organizer, and had very little computation. I think there is a huge difference between actual Microsoft Office Excel and the clones like Google Sheets, which is a pretty mediocre copy when it comes to capacity.