I'm sorry to ask for justification for a change as an outsider from the project, but that does bring up the question of whether getting rid of direct calls is as urgent as to justify such a large regression. Are there any plans to keep direct calls around for longer in Proton? Especially considering Proton's target platform, which has a relatively weak, battery powered CPU, I don't see how even adding 5% overhead to very hot vulkan functions would be acceptable.
It's possible to improve the winevulkan thunks a bit, to mend the gap between the current direct calls and the new path. It's still not going to be anywhere close to unix native performance but I think it makes the change acceptable.
Regardless of whether the calls are done directly to the unix thunk or through the unix call dispatcher, the ABI conversion we have to do will always hurt these very hot functions a lot.
There are ideas floating around to workaround this, for Proton specifically, hopefully reaching something that would be close to unix native, but it's not going to be easy and it's probably going to require doing things unlikely to be upstreamable. It's also still hypothetical at the current time.
In the meantime, we will consider anything that seems useful for Proton, including going back to direct calls if it feels necessary to avoid a performance regression.