It is such a clean base architecture on one hand, on the other hand it simply abstracts away lots of stuff and calls it a day. Most relevant to me: any and all accessibility-related APIs.
I am quite convinced that Wayland won't ever really take off in its current form. Maybe someone integrates Wayland into a big package with all relevant APIs fixed and documented, and we get rid of the compositor abstraction mess we have.
Also, I'm not really convinced that creating Wayland as a totally new development was a good idea. As we all know, greenfield projects to replace existing technology are almost always doomed. But in X's case, this may really be the best solution, I don't know and I'd lean to give credibility here to the (definitely incredibly talented) developers behing Wayland.
Just one of lots of examples what's not so polished with wayland¹: https://community.kde.org/Plasma/Wayland_Showstoppers
There are lots of other links (see my comment history). I admit I am a bit sour.
¹ yes we're getting there, but only when all compositors and DEs have implemented their own APIs and solutions to the common problems wayland simply ignores