I had some deep look into webmx implementation. As I understand it basically the graph of all cells is automatically corresponding to how the html dom is structured. I think this would allow a variatio of webmx that could generate the entire html in the server and then just send updates of fragments of html. In a way similar to htmx; but instead of htmx there just would be clojure code that generates dom update Fragments. Just some thought....
That actually is how web/mx apps work. We create a seed HTML that gets hydrate with the HTML of a full landing page, after which web/mx just tweaks the DOM as necessary after each event. In the private qooxdoo library, it works the same, but the initial landing page is received from the server as a large wodge of JS to build the DOM client side, after which events are relayed to the server so it can emit new JS to tweak the DOM. So you are on the right track!