What in particular about the WebVulkan memory model do people feel is incompatible with the web?
I'd really rather like to know that too.
There are two main benefits to the (somewhat complex) memory model that Vulkan exposes:
It allows you to figure out how much of what type of memory there is and adjust your usage. This is a really, really, really big thing. There's a ridiculous amount of memory non-use happening with many high end applications I deploy, even though selecting quality for this case (texture sizes for instance, or different model LODs) would be really trivial. But because memory is unknown, and because mobiles have teensy memories they don't like at all being overstepped (<ooops pages> everywhere).
Allocate things in the most appropriate memory region for your use (this can make a huge difference in speed in some situations)
These two things are why Vulkan added them. Because it's become a huge headache for native developers to address exactly these issues. It's exactly the reason why they should get to the Web, because it gives native an advantage, an advantage we don't want to let shear too wide open.