Not the OP, but taking a look at the DOM, it's rendering each "window" as a canvas. So you have a main window canvas, then canvases for popup menus, message boxes etc. All canvases are using a 2d context, and there are many offscreen canvases created for each interaction for each widget which I assume are then composited to the window canvases.

It's an amazing project! If the OP is here, do you have any plans for accessibility to screen readers? I suppose it's a similar problem to that with Flutter.

I also wander if there is a possibility to combine your wxWidget port with the work going into WASM Python...

Edit:

Taking a closer looks at how the wxWidget port works, wxWidgets has multiple backends using native widgets on each platform. They also have wxUniversal which draws each widget itself to a frame buffer, this port uses that as a base to draw the ui to html canvases.