I've started building a spreadsheet program where all the tables are built from a single commit log, so there is a diffable history https://vimeo.com/727150523

This is cool! I am curious what kinds of features a single commit log might unlock. (Aside from the diff view you’ve already shown, of course.)

Good question. I think diffing that understands a table at the cell level is probably the majority of the benefit. However the fact that the tables are not the source of truth but the log below them means you can have different tables for different uses, potentially needing less joins. Another area I'm looking at is can you derive tables from more than one commit log and use this as a potential way to collaborate.

If you're interested in history for tabular data, you might be interested in "TACO: Visualizing Changes in Tables Over Time": https://jku-vds-lab.at/publications/2017_infovis_taco/