A separate process has an overhead over just using a separate thread. If you put each individual tab into an individual process, you will have this overhead for each individual tab, meaning that you're much more limited in the total number of tabs that you can have, no matter what those tabs have loaded.

Is that overhead inherently high, or does it just happen to be high given browsers' current implementations? My understanding is that the inherent overhead of an OS process is fairly low, and that it would be possible to implement process-per-tab in an efficient way, but I don't know enough about the low-level details of how that works.