Dissecting the Apple M1 GPU, the End

I wouldn't call Vulkan 'vendor neutral', but instead a collection of more-or-less adhoc design decisions to make Vulkan 1.0 work on specific GPUs that didn't match the AMD architecture (which Mantle was designed for, and then when Vulkan was derived from Mantle it had to add all the bells and whistles to extend the Mantle programming model for other GPU architectures, especially mobile GPUs).

From that PoV, at some specific point in time Vulkan might indeed not have been a good match for Apple hardware and might have required extensions to make certain features work efficiently (and those extensions might then later be rolled into core Vulkan). FWIW though, since somewhat recently Vulkan and Metal have started to 'harmonize' a lot (Vulkan via 'new' features like dynamic rendering - which is pretty much a 1:1 copy of the Metal render pass model, and from the other direction, Metal4 adopted some features from Vulkan/D3D12).

Metal was developed 12 years ago for the iPhone 5s. Apple was clear (and probably right) that a 1st party system targeting low power devices was more suitable. That became much more of a 'myth' 6 years latter on high power laptops people wanted to play games on.