The statistic that really matters: desolation!
I questioned Hypergrid Business' measure of landmass as a means for how good a grid is. So I looked at their October's "most active" grids stats from HypergridBusiness, and worked out some spreadsheet magic. The difference is, that I look at the number of actual logged in users (concurrency) in relation to the number of regions the grid has (single size equivalents). And this gives a much more usable metric as shown on the graph.
For OsGrid for instance, this means for every user that is logged in, there exist 201 regions on the grid; 200 without people, and 1 with a user.
If we want to bump into a another live person on Wolf Territories, we have to travel through 2611 regions!
It is no surprise the microgrids do well here. They make the grid no larger than is reasonable for the grid's concurrency.
Best in this graph is of course Second Life, with 0.6 regions per logged in user. That means every region in SL, has on average about 2 users in it.
Data file available at https://lotix.nl/posts/2023-11-03-grid-desolation.ods
(you can repeat the test yourself)
NB. data measured at 19:30 UTC, usually when most OpenSimmers are online
NB2. the concurrency reported by Alternate Metaverse is most likely incorrect/fraudulent; correcting this to 60 (like when there's an event/party) yields a desolation of 167 regions per logged in user.