The speed problems in OS X's kernel doesn't appear because of the fact that its monolithic vs. micro, its because memory allocation is painfully painfully slow.

This has been my observation in my optimization work in my own (unreleased) app, and also the recent OS X vs. Windows stdlib performance tests.

This more than anything I hope gets fixed because I think it has the potential to speed up high level frameworks like cocoa that allocate and release lots of auto-released objects.