Moving to Xegl would have been steps backwards until appropriate approaches for the problems I mentioned could be found. (How would existing stereo 3D GLX applications work? Very important to the people that use them. Video was a mess, too, what about multi-screens...) David was rather flippant about these problems at that time, but they were real roadblocks.
AIGLX provided a simpler route that didn't lose functionality in the meantime and didn't require writing a new driver. It wasn't just Red Hat / Nvidia bullying their way through here (your premise up the thread) - it gathered momentum as it was (at that time) the simpler way forward that didn't break anything. I think some of the Mesa drivers were first to implement AIGLX.
I'll agree Xegl did show promise in providing a single simple approach for driver development though.
So, given the nature of open source - if Xegl was superior, why didn't someone keep working on it? Why aren't we discussing Mir and Xegl instead of Mir and Wayland?