https://bugs.winehq.org/show_bug.cgi?id=56380
--- Comment #12 from Rémi Bernon rbernon@codeweavers.com --- Thanks for testing! I may need a bit more time to consider the implications of the fix, before sending it upstream. It may also be related to how DXVK used vulkan surfaces as I couldn't reproduce it myself.
For the record, the issue is that this change also lost some vulkan surface invalidation that winex11 was doing when new surfaces were created on a window which had some surfaces already. Later vkGetPhysicalDeviceSurfaceCapabilitiesKHR calls on the invalidated surface succeeded when they previously returned VK_ERROR_SURFACE_LOST_KHR.
This is not straightforward to fix however because of how we use vulkan to implement D3D. I believe, though that will perhaps need a better confirmation, that Windows allows concurrent APIs to be used on the same window, which will break if we only allow a single vulkan surface at a time.