OpemSubDiv in 2.79 was less complete in its implementation, but it had both CPU multi-thread and GPU acceleration options. I’m not sure why re-implementing it would be a difficult thing to do.
Having a priority on performance for 2.81 would be a critical factor if the BF wants to stand by their tagline of the fun being back in 3D, how fun would it be to pose a character or edit a model at 1-3 frames per second?
It would also be critical because otherwise, Ton risks losing his license to excoriate Autodesk on their business practices, as he wouldn’t have anything that’s objectively better than what the ‘evil’ corporations are putting out.
Fortunately, this is something that is very much in reach as 2.8 is almost there already. Moving the view in editmode is already fast and it’s mainly geometry transforms that need a boost. OpenSubDiv acceleration should take care of most posing/deformation problems. The BF has also demonstrated for the 2.8 cycle that they are willing to let go of some of the old ideological points that held Blender back. Let’s not have the old FOSS ideology problem slam the brakes on any chance of Blender being a serious competitor to Maya, as this would likely be the most serious threat facing 2.8x at this point.