cool. will check this out. i think it's an interesting approach and allows all sorts of very low rent interactivity as long as you don't need super high throughput or expect lots of contention.
I purposefully replicate data and queries only, so to change the shown data for everyone you have to run a query, so after the import process (which can be slow for large files), queries are replicated fast enough IMO because it's just a small string sent over WebRTC, so in theory it should be faster to run a query over a big dataset and see its result appear for all peers than a centralized app like Google Sheets, but I haven't benchmarked it yet.
