Even greater would be M$ open sourcing DX but that will not happen in the next few years.
I would assume that performance would be low-hanging fruit. OpenGL and DirectX work quite differently, so translating the calls of one to the other should involve quite a bit of work/CPU cycles. Vulkan, being closer to metal, should be a more straightforward translation target.
Example: HLSL-to-GLSL is likely a difficult problem, where HLSL-to-SPIR-V is probably much simpler.
> Who cares about DX9 nowadays?
Lots of games still support it. In addition this layer could be used for future work in DX11/DX12, instead of the (per my above argument) more complex OpenGL translator.
Shouldn't be harder than what Wine is already doing. Open sourcing can surely help in some ways. MS for example already open sourced HLSL parser, so translation tools can benefit from it.
DX11 support in Wine is progressing quite well.