If you have a class full of "good" students, it's a lot easier to work and teach, than with a class where 1/3 is almost illiterate, 1/3 doesn't care at all, and only 1/3 actually wants to learn.
I attended a talk about this topic that used the PISA dataset to show this (well it showed that in countries with segregated schools the performance of students had fatter tails on both sides of the mean iirc). I think the data should be available online somewhere.
Yes. It appears that the more stratified educational systems are, the more variance in outcomes there is. e.g. [1] Which makes intuitive sense. While there's an intellectual argument for everyone benefiting from diversity of a student body, I'm not sure I'd expect to see a benefit in strictly measurable educational outcomes at least.
And it's fairly obvious some of the ways that motivated and high achievement students could benefit from more customized and self-paced study while it's also obvious why a group of students lumped together as low achievers will do less well.
I went to a highscool where there were about 20-30 International Baccalaureate Diploma candidates and about 1200 students in regular, honors, and a handful of AP classes. My junior and senior year I was in all IB classes, but freshman and junior year I had to take a handful of the regular, honors, and an AP class and they were all absurdly slow and easy to be successful in to the point that I would entirely disengage. Had I not been in such a rigorous program I think I would have almost certainly made worse choices in relation to my education and life outside of school.