You know, I was wondering this myself for a while now. But then I did some tests on some games and saw that they do work... sometimes. They are basecally process-name based override hacks that intrude on executable's memory when GPU's userspace driver parts load. Sometimes they just fail to do anything, other times they actively interfere with what process is doing, and very rarely they might actually help... very rarely. The only things that may be usefull are antialiasing override and "morphological" postprocessing antialiasing. But pretty much in all cases they make everything look worse, don't even load or hang. Disabling vsync is usually a stupid endevour because game may use it to sync not just the visual but all its input/output threads. What is problematic is "buffer bloat"-like issue of pre-rendering more than 1 frame (previously controlled by "flip queue") which multiplies input lag as if you had 1/2 of native or worse fps while showing you high fps.
I would rather liked if they would spend their time to make 100% compliant and 100% open-sourced OpenGL/Vulkan/OpenCL/DX Mesa/LLVM-based framework and used "profiles" to enable or disable their hidden hacks. You know, the ones that they make for every new high budget game or every new GPU released which fails horrribly on launch date. Instead they continue to dumb UI down and rename it with weird names... Why the hell you even need to advertise interface to your driver's hacks ?!
And maybe make antialiasing algothithm that actually works, move away from rasterization or at least push for bigger chips for which 4k resolution and >100 fps is no issue.
