https://www.syntax-k.de/projekte/teapot/
In short cell address are normalized @(1,2,3) instead of A1 or r1c1. real references so address rewriting hacks($A$1) are not needed. formula references so you can use a single master formula, and clocked expressions which allow circular dependencies/simulation.
Probably a little too different for casual use but worth taking a look at, if nothing else to challenge your ideas of what a spreadsheet has to be.
While looking up the website I found a rewrite in rust, which is cool I guess, someone is keeping the dream alive, I will leave a link to that as well.
You can also literally run Lotus 123 if you want. Someone has binaries to make it work on linux. or under dosemu
Edit: Quite painless! Opened some test xlsx files without issue. Did get a stack trace on a very complicated one, so when I have time I'll try and dig in deeper. Added a doc to the wiki in case it's helpful to other: https://github.com/andmarti1424/sc-im/wiki/Building-sc%E2%80...
"But... visidata is not a spreadsheet"
I know, that's what makes it so weird.
On contemplation, I think I grew dissatisfied with the normal spreadsheet data model, I wanted something bettered structured than the "it's a big bag of cells" that spreadsheets present, I wanted row security. The best I found was the relational database. I currently use a local postgres db for most things I would have used a spreadsheet for. The interfaces sort of suck in comparison but at least I have sane data structures.