Yep, I suspect the goal is to reduce support footprint/burden to the absolute minimum possible, which is also why they don't maintain their legacy virtualization/translation layers indefinitely (Classic mode, Rosetta 1 (PPC → x86), Rosetta 2 (x86 → ARM)). The benefit is being able to make major changes, refactors, etc with relatively little delay and encumbrance.

This applies on both the software and hardware layers. In the past, CPU suppliers dragging their feet and/or being unable to deliver on promises were the chief motivation for switching architectures, and now that they're their own supplier they can banish that issue entirely so long as they're not beholden to any kind of external behavioral contracts or expectations.

If the OS teams need some new capability or discover substantial benefit from redesigning some part of the SoC, the SoC/hardware team can just implement it and not have to care about breaking anything. That's a major competitive advantage that practically nobody else in the computer or mobile device industries has.

> they used to let you run Windows!

They let you run Windows when they were losing in the market and needed a path to migration for users of the winning platform. Once they started winning, they stopped.

Conversely, Microsoft refused to interoperate with Linux. Until Linux became the clear winner in the datacenter space. Now they have WSL and actually work pretty well as a cross-platform development environment.