You're right to point out that I'm in the optimistic camp on this issue. I think it's both do-able over the course of several years of refinements, and necessary given that I think we're not done yet with innovation in interaction methods. We've been on a plateau for a while with iDevices and Macs. Meanwhile, foldable screens, touch surfaces replacing keyboards, and gaze tracking are on the horizon, and those will require some fundamental UX work to be done across both iOS and MacOS, making them even more similar than before. Those techs are Even without new interaction methods, this kind of work is not an endeavor I think most companies can undertake, especially not Microsoft. Their failure doesn't reflect on the opportunity or possibility, because they really lack the deep bench of skill and company culture of perfectionism to get the UX right, and they don't have the market leverage or third party developer culture to push developers toward new things (see win32, uwp, etc). Win8 was a half-hearted springboard reskin that was only surface deep, and in no way a re-think of combining touch+mouse, let alone meaningful guidance and tools for developers on how to make apps span both. Users rightly rejected it and Win10 went back to a more incremental approach that's still mired in weird reskin issues, and touch is largely an afterthought on Windows except in specific apps.
Re: Mac App Store, it isn't actually a Marzipan app, though it shares some of the questionable UX choices. I'm particularly not fond of the full page modal with 'Done' button rather than the more traditional back/forward design. Totally agree with you on this and the Marzipan quality issues.
Discord is sort of representative of an aspect of the problem a combined UIKit is looking to solve - devs are using things like Electron to get cross-platform support but creating crappy anachronistic non-native experiences in the process. If devs can build for UIKit and get some platform-specific graceful enhancements built-in (as they did going from iPhone -> iPad), that changes the calculus on building with Electron or whatever.
There's a lot of implementation details regarding Marzipan and UIKit but - saying this as a designer who's had to do a fair share of mixed mobile-desktop work - I don't see any fundamental blockers to the approach that would prevent pro apps on the Mac getting their interaction density on mouse and touch apps getting their lower interaction density touch targets. Just a lot of work and decisions to make around how to progressively enhance and offer platform-specific flexibility. But that's just my take, I can see why others might disagree.
We'll see this June. Apparently some more fundamental UX changes were delayed last year to focus on quality, so this year is likely a big year.