Just as an example: JSON data - Including a Swift serializer adds code size, but bridging to JS is a bit less comfortable. And for each similar feature we will need to find a good compromise. We need to be careful not to create a "parallel universe" of Swift, where everything works completely differently from "normal Swift" - but at the same time things like 'import Foundation' won't fly on the web.
And, of course, things like HMR or "module splitting" and other more dynamic features are much, much tougher to do with a compiled language and Wasm. So, trying to reach "DX-parity" with JS frameworks can be a very pricy endeavor.
I think Swift adoption in general is also a ... factor. I personally think it is way more ergonomic to write than most languages, especially for UI and app-level stuff, while bringing a lot to the table. But as - we all know - it is still hasn't grown to its potential outside of Apple-land.
Embedded Swift is a moving target, but I am confident it delivers (it's evolving rapidly). Also, Wasm is a moving target to a degree.
And then there is just a ton to do, none of which a "great challenge", but just the grind of moving towards a "fully-featured UI framework"...