How hard would it be to merge MobileFrontEnd, or parts of it, into core?

We are currently in the process of packaging MobileFrontend's JS into a standalone component library. When we're done with this, the library will live outside MobileFrontend (similar to OOjs UI), any remaining code in MobileFrontend will be refactored so its compatible with core (and any relevant upstreaming), but the majority of the code will end up in Minerva (which makes sense - just like VisualEditor houses VisualEditor-OOjs UI code). Follow T171000 for updates!

The code that remains in MobileFrontend will be limited to special pages and mobile formatting. I'd be keen to see the special pages in MobileFrontend upstreamed to core and rewritten using core's libraries. This includes https://en.m.wikipedia.org/wiki/Special:MobileLanguages/Spain and https://en.m.wikipedia.org/wiki/Special:Nearby

Would that be worthwhile in terms of maintenance? Development of new features?

Long term it's hard to say. Minerva is pretty sandboxed from core and is essentially a mixture of mobile-specific components that don't make sense in a desktop context. I see skins as separate experiences and thus free to use their own code (just like many skins make use of bootstrap).
Time will tell, but I think it's too early to answer this question fully. I can say with all certainty though that MobileFrontend is converging towards a much simpler extension that simply provides an HTML formatter for mobile content and the capability to support a different skin on a mobile domain.