It is unfortunately nigh impossible to reimplement the web with its gigantic scope.
A realistic scenario could be something along the lines of: get content onboard by securing some buy-in from some heavyweights, for a reduced feature set. Maybe convince a incumbent browser vendor or two to also provide a safer kernel for that content profile. Then targeting that feature set with the new implementation, and supporting a compatibility mode fallback to a legacy rendering engine for other web content. (VM-based or remote cloud based sandboxed rendering for the fallback?)
But the above would of course have a much better chance of taking off if there was a promising start of an implementation first, so it's not necessarily a reason to wait for the big boys before working on it.
(I wonder if Servo could be a starting point?)
Firefox is Tor Browser which is the most attractive target by far for 0 days. 0 days are often not targeted to a single individual anymore but are commonly used "in the wild" against mass targets, e.g. everyone who browses a political dissident website.
We're reduced to reasoning like this because we've in browser security apathy.
If only there was a browser that wasn't implemented by juggling chainsaws like this... It's ridiculous that we have been unable to just fire all the these apps for sane reimplementation. Feeding hostile internet content to millions of lines of unsafe code day in, day out on billions of user devices would sound like a ridiculous idea if it were invented today.