In the PowerPC days, Apple had trouble convincing people to switch to Mac because Windows being so dominant, everybody was afraid they might need Windows for something and that would make a PowerPC Mac an expensive mistake. The Intel switch alleviated many fears because in the worst case and Mac OS didn't work out, they could just install Windows. Bootcamp and virtualization provided additional options.
Apple has always seen Mac and iOS as different markets. Mac is still a small market compared to Windows. They are still trying to grow which means still trying to convince Windows users to switch and the safety net of Intel is still useful. (Windows RT isn't a realistic option.) I personally haven't seen iPhone users wishing for a big laptop or desktop that runs their same apps.
Windows was already on ARM with Windows RT, for which I believe Microsoft did the work of porting UEFI and standardised bridge (PCI) technologies to ARM.
While speculating: Apple could conspire with Microsoft to standardise these technologies together to lock up their ecosystems against the Linux / Android monster :-)
The reason Windows on Mac is useful to people is because people have may programs that do not have acceptable equivalents on Mac OS. These apps are usually legacy programs that probably fill special niches, and the cost of "modernizing" is usually not justified. This effort would be required to port these apps to Windows ARM, which last time around also required porting to UWP which Microsoft is still pushing, but not easy to actually do for a lot of code bases.
A new Microsoft irony kicks in here is that if they are successful in convincing the huge Windows ecosystem to migrate to UWP away from Win32, et. al, they may actually kill their lock-in advantage. A serious rewrite at this stage may also invite Mac, Linux, iOS, Android ports. In this case, Apple no longer needs to care about Intel Windows and this bullet point becomes less compelling and maybe they could reconsider.
(But I'll freely admit that me being able to run games at sub-30fps on my MacBook Air is probably not Apple's intention.)
It wouldn't work for side loading unless doing it via Visual Studio.
Say, does anyone know who owns the IP for FX!32 nowadays?
Remember they were called Powerbooks back then? I miss that name.
People are concerned that they have some mission-critical, unreplaceable app that only runs on Windows, that has no acceptable Mac analog (doesn't exist, doesn't have feature parity, data formats are incompatible).
Windows on ARM doesn't solve anything because almost no Windows apps that people care about were ever ported to Windows RT. (It was probably more likely there was a Mac port.)
And for a lot of the older Mac developers, iOS was kind of like another migration. And iOS has gone from armv6 to armv7 to arm64, plus the x86 and x86-64 simulator targets. Not to mention that there is now an LLVM Bitcode requirement for Apple TV and watchOS.
Apple and its developer community has a lot of experience with architecture migration. Each transition built on the experience of the previous and got smoother each time. Apple has been very good insulating their frameworks and tools from the architecture, and the Apple developer community has gotten very good at following Apple's guidelines to minimize disruption since there have been so many of these transitions.