Try giving foobar2000 a spin or checking out some screenshots to get an idea for what I mean. Unlike Chromium, the browser should be extendable, at runtime, via a component system and editable layout tree. Runtimes, backends, common libs, fonts etc should be managed using a repository system.
Chromium requires recompilation (and know-how) in order to make any big change. Comparatively, a user should be able to completely reskin their browser, switch JavaScript engines, etc. on a per-domain, per-group or per-tab basis in seconds without restarting the browser.
Integration is relatively easy to solve. I did not use foobar2000 for a long time; I can compare instead with Emacs, where various packages are aware of various other packages and will integrate if a counterpart is present, but won't require that.
Security is much harder; the need to isolate pages means relatively smaller customizable API surface, and, most importantly, greater trust to the components. It's much like web extensions now: you have to think well before giving an extension access to your web mail or online banking web pages.
XUL was an attempt to build a very flexible architecture, and it was great, but the (lack of) security killed it.
But to your point the core browsing experience would remain the same. Would be cool if they opened that up to variability though.