Because greenfield parallel rewrites of giant production codebases have such a fantastic track record, that we might as well add "in a novel programming language we're inventing while we're at it" just to make it interesting?
Security is a huge concern for something like a web browser, which is a thing that keeps on increasing more and more in complexity. So far it seems Java is dead in the desktop space, and .NET has a very minimal impact, so it seems a compiled language is the way to go.
And Rust's excellent support for multithreading is also a big boon since now multicore CPUs are a standard and keep on getting more and more cores.
Overall the idea of building a very safe and performant browser seems like a great one.
I think demise by drastic jump stories work because the examples are few and notable. Most projects accumulate experts in moving around the wrong optimum who then block plans of getting free that aren't too conservative to possibly work.