It's not really about the OS. Vulkan should be equal or superior to DX11 in all cases even with this shader compiler limitation. Even if the shaders aren't optimized, Vulkan is intrinsically faster than DX11, therefore in the worst cases I'd expect framerate to be in the same ballpark than DX11. In the best cases ( a game that has simple, pre-optimized shaders or that isn't fill-rate limited ) it would be significantly faster than DX11.
Chill Vulkan and DX12 are vastly superior to DX11 and OGL. Even with that 30% loss, you still probably had a 30% gain in the first place just because of the better api / less overhead. What it truely means is that unoptimized Vulkan is preventing your app for reaching its top potential, that's all. For example:
OGL/DX11: 50 fps
Optimized Vulkan/DX12: 80 fps
Unoptimized Vulkan: 40-60 fps maybe ?
That's an hypothetical scenario. The only real number we have right now is based on a single sample, the Talos Principle. Each game would be different, and I'm sure there will be games which won't be affected that much by this problem. On the other end of the spectrum, I would expect us to be close to the worst case, since we have some uber-shaders that generate thousands of lines of shader code. Those aren't even manually generated ( pre-optimized by us ) in the first place, they come from our own material compiler, so we rely heavily on the shader compiler to optimize it.
I was debating all of this with Keith the other day, and we reached the conclusion that we'll probably try a simple comparison benchmark ( a full-screen quad with a super complex unoptimized shader ) before we take any final decision ( this should only take a day or two to implement ). That way we can figure out if the problem is as bad as we think or if we over-estimated it.