That future possibiilty reminds me of https://en.wikipedia.org/wiki/Singularity_(operating_system) - where process/address-space isolation was replaced with fine-grained static verification of high-level code (presumably not the first experiment in this area).
Indeed: that and many other things are prior art in this space. And there is a lot of prior art for what we're working on—this is not meant as an academic research project! :)
Yes, one of the answers I want to give any time someone asks "why will WASM succeed when the JVM didn't" is that there is 25 years more experience and research to draw upon.
And yet bounds checking access validation was left out of the design, something that most of previous research projects took care to taint as unsafe packages when present.