I can't figure that a DX12 switch is much of an option because many supporters might not want to use Windows10.
Given all the vendors and developers going to Vulkan, perhaps Microsoft will cave and bring DirectX12 to Windows7 and Windows8, but that's a way crazier thing to bet on than the much more likely scenario that some vendor is going to solve the Vulkan shader compile issues and make their knowledge on that openly available.
I can only see the question being sticking with DX11, or switching to Vulkan now while it's easier to do earlier on, which offers more for the future.
The devs can answer better than I, but it's my understanding that adding alternate capability of DX11 to a DX9 or DX10 game is much easier as they are incremental changes while DX12 and Vulkan are much different and require you to program at a lower level and make big changes to how the programming for your game's renderer sends those instructions.
But I'm also on the understanding that that rendering pipeline for Vulkan and DX12 is similar, while Vulkan supports the same hardware as DX11, so adding an alternate DX12 renderer to someone that started on Vulkan may be easier than adding DX12 support when starting with a rendering pipeline built for DX11. With the exception of the shader language being different, but it's 95% the same and I believe you can pretty much automatically convert from one language to the others? At least you could with DX9 HLSL to GLSL.
That does mean it's going to be work to take the current DirectX11 rendering to use Vulkan instead. But all that rendering isn't finished anyway, so perhaps now is the time and it would make going to DX12 easier later on? Or to save the most time, but be limited both in capability and to Windows(no Linux or Mac down the road), just stick with DirectX11. Though it also can mean that Vulkan eventually makes things easier down the road to get the performance and capabilities that are sought (assuming shader fill issue being resolved, which seems like the only big thing here while other vendors are showing huge performance increases, especially with AMD GPUs).
I think a lot of people are confusing Vulkan as being a snazzy new bleeding edge thing that's unstable and might not stick around, but it's not. It's really like a more mature version of Mantle, which is what, two years old? And that's going to mature more. And it supports all of the same exact hardware DX11 supports, if I'm not mistaken. It's a new way of doing graphics programming to have the programming fit the way the hardware works, while DX11 and earlier were all based on like 7+ year old GPU architectures. Both Vulkan and DX12 are things that were talked about for a while, and gone through a lot of development to catch up to match the capability of the incremental graphics APIs.
They're the beginnings of the new and hardware has been long since out to support this way of programming, while DX11 was the last crutch of the old.