I briefly looked into using vdom for the editor preview and it was a mess. Basically we spit out HTML from our markdown pipeline. That HTML would have to be parsed before it could be diffed against the dom, in other words every keystroke would have to parse all the HTML first!
I punted on it as we plan on moving to CommonMark, and I was hoping we could have marked emit vdom nodes instead of HTML. Then we could skip the parsing phase altogether.