Here is a question for you, if MonoGame kept its existing graphics API and added Vulkan/Metal/DX12 support, what benefit would the devs / end users see?
Zilch. Unless being thrust into the driver-hell that Vulkan is apparently trapped in for perpetuity counts as a beneficial experience. Probably be able to get the same experience 2 years from now.
DirectX 11 and OpenGL 4 I think support it (I might be wrong on this)
GL 4.3+, anything less lacks SSBOs (roughly equiv. to DX11's UAV) and without those compute is just plain silly.
Time saved by not loosing time trying to recreate and optimize an effect that would be more efficient with compute shaders
I've been working with compute since the dawn of CUDA - it's anything but a magic bullet. It's more often a PITA of histos, fenwicks, and chokepoints that barely wins out over OpenMP in a narrow range.
You still have to write a CPU fallback and that's going to remain the case for many more years unless you have some incredibly narrow target (like console XYZ). nV has also only recently started getting on async-compute, so you really can't do as much as you think, the greater majority of even recent hardware lacks async and that hurts.
You need an incredibly narrow problem, like GPU particles, for it to actually be a no-brainer. It's much more powerful at design-time than runtime.
spending more time on the game whereas spending it on "how to do this or this without ComputeShaders".
Compute in both GL and DX is incredibly trivial API-wise. OpenCL and CUDA are trickier but they still take far less than 8 hours to get rolling with. All of the hard is in making algorithms fit that much parallelism without requiring so many prefixes/histos/syncs that regular threading ends up faster.
If you actually need it (and can actually use it) then it shouldn't be a problem.