Assuming that the effort is focused on making frameworks that simplify macOS development for folks trained as iOS developers, then this will have a large impact on native Mac development.
As a user I personally prefer more powerful software and a more financially motivated developer; even when it makes buying more bespoke.
Last I checked, I had to assemble the above from disparate services, investigating each, integrating them, and paying for each, which doesn't seem like a productive use of my time or money.
Assembling from disparate sources will have more of an upfront cost but cost less per purchase. So it depends entirely on your sales volume.
Regarding your second point, interesting, seems like MAS is good to get started until the app reaches a certain volume.
Apple could unify UIKit and AppKit so you could more easily share parts of widgets and the backend code, but you'd still have to design the general user interface separately.
Two possilities here:
They dust off NextStep’s YellowBook support (Cocoa for Windows) and update it to support UIKit.
-or-
What if the parameters of Marzipan is not a compatibility API, but an effort to package the iOS Simulator and your x86 simulator builds transparently for the end user. Then port the iOS Simulator to Windows.
Halo Effect 2.0
And several Apple software like Safari 5 for Windows and iTunes already ship with macOS UI libraries ported to Windows. So in near future iOS apps may run on macOS and Windows - quasi true universal apps (not the UWP-crap)
> several Apple software like Safari 5 for Windows and iTunes already ship with macOS UI libraries ported to Windows.
Actually, the situation is kind of the opposite: Safari and iTunes eschew the use of macOS's UI libraries. Safari, in particular, is actually just WebKit (which has no "UI") wrapped up in a Windows-style chrome to appear native. iTunes itself, even on macOS, is some sort of weird C++ hybrid abstracted API that's not Cocoa (which is why it looks so "off" on macOS). This, along with a version of Foundation/CoreFoundation, allows it to work on both platforms–though not very well, I'll admit.
Safari 5.1 (Win32 version, discontinued) comes with several MacOSX user mode libraries ported to Win32 DLLs.
Older iTunes and QuickTime (like 5 years ago) came with MaxOSX libraries ported to Windows too - at least from what I recall.
One day there might be an opportunity for a web-focused phone OS (what Firefox OS was trying to be, essentially), but... not quite yet.
Compared to a native app web apps are bloated, slow, have foreign UIs, and are power hungry. On a desktop machine, it probably doesn't matter, but for anything battery powered it does. I'll choose a native app any day of the week over a web app.
To be honest, I don't see a persuasive argument. In order for this to be worth the extra development time the company needs to have a big marketshare on both platforms - MS failed because they had zero footprint on mobile, and I think Apple will fail because they have a relatively small footprint on desktop. As a third party developer, is it really worth your time to make a native macOS app when the web is already very capable? In some cases I'm sure it will, but in the vast majority of cases you're going to need to make a web version anyway (for the huge number of Windows desktops out there), so a native macOS app just wouldn't be worth it.