Browsers are catching up of course but... Who would thought that the thing for reading pages will need access to your file system or make screenshots...
#3. UI engine must have natural and effective means to use native code - to be embeddable and extendable by hosting application.
In native app world you don't need WebAssembler for example - you can run native code you need natively.
It does! Most browsers are written in "native code" and use it to do layout. When you use an <input /> field on a website, you get a natively rendered UI widget with native OS behaviour!
Consider that you need add <word>...</word> element that will implement natively MS Word editing area. This will allow you to define MS Word editing chrome in HTML/CSS but keep core editing functionality native and effective.
Consider implementation of <tabs> native element/behavior: https://github.com/c-smile/sciter-sdk/blob/master/include/be...
This is a demo of how native code can extend DOM - adding native "drivers" of app specific DOM elements. Yet there are other examples of native behaviors: https://github.com/c-smile/sciter-sdk/tree/master/include/be...
Essentially native behavior can do anything that built-in elements can do, even drawing on the same surface that the rest of DOM is using.
Good luck with achieving anything like that in browsers, or in Electron, or with WebAsm or whatever.
What should you do if you need your app to run on W2000 for example (IE 6.0 anyone?) ?
Sometimes people do need that: https://sciter.com/necromancing-sciter-on-windows-2000/
One of the main advantages of shipping your own browser like Electron or SciterJS is that it's a fixed, stable target where you can be a lot more confident about browser compatibility because you're only supporting a single browser + version.