I'm not sure if this is a good idea. Components and extensions have a ton of overlap in functionality even if the DSC pattern bars you from using components from extension-like functions. Having both systems in parallel feels like a complexity and maintenance issue.

I agree that the two-implementation solution is messy (See: nanoui v. tgui, among probably numerous other examples), but there actually isn't that much that uses components right now.
I much more often see shiny things that wecould have, but for lacking extensions, than I do shiny things that we'regetting because we have components.

I would want this to come with a commitment to transition existing component use to extensions, which I think defeats the "I much more often see shiny things that we could have" sentiment which, I assume, carries the caveat of still wanting the shiny things from components without the effort of translation. I agree with MistakeNot4892 - both is a recipe for overhead, interoperability issues, and dev headaches.

I'm biased in favor of extensions because they're what I use already elsewhere. I've found DSC a tad less approachable, largely due to the rote use of its own signals system creating a lot of async funnies and signal comprehension instead of direct calls.

Buuuut it is what currently exists. Translating extensions to it is just as possible.