One aspect of the introduction and transition to Wayland that's being ignored is the impact on distributed computing projects. There's are many people who dedicate their computer cpu and gpu resources to projects via BOINC and World Community Grid who use closed source proprietary drivers due to project use of CUDA or OpenCL. For people doing gpu work for BOINC or World Community Grid the open source drivers are not an option. As distributions start implementing Wayland these users will be forced to make a difficult choice - continue to run an obsolete release version, switch to windows, or give up on gpu computing. This would be a hard blow to the scientific research being conducted by the distributed computing projects and significantly slow progress being made in many fields of research.
I'm pretty sure CUDA/OpenCL software mostly rely on the kernel bits of the drivers and low-level libraries that don't need a graphic server to be loaded by the kernel. The Tesla thing from Nvidia is not mostly run on windows, so for pure GPU computing, I guess it's the same. And then, we're not going to kill x.org this soon. Even at 1.4 version, Wayland is still considered a work-in-progress architecture (despite the stable ABI), so seeing projects having some trouble with it (see "classic" DEs like GNOME and KDE, and now the Nvidia driver) is not something "alarming". Time will see finalization of most of the work. And so GPU computing as well.
As a side note : it should be interesting to see OpenCL benchmark between cloed and open source drivers (at least for AMD, until nouveau devs are able to unleash Fermi+ frequencies)
As soon as a wayland binary blob comes over the switch will be instant!