That might be reasonable for a general purpose computer if we were talking about something like a Parallel Inference Machine running KL1 software on a KL0 kernel. But I think conflating Apple's products with that level of foundational engineering is highly disingenuous. They're not exactly trundling into the dark woods of exotic hardware and reinventing the bridge between human and computer. It's an ARM computer running a Unix clone. Apple's engineers aren't mapping every codepath and counting every micro-op, Darwin contains extensive amounts of third-party code.
Hardware and software have to interface at some point. When the people designing the hardware work at the company designing the software it's not unreasonable for them to come to some shared understanding of that interface which may not be standard, portable, or even publicly documented, and certainly not one that is stable.
This isn't incompatible with allowing users to install their own software. There just isn't an obligation on the original designers to make sure that software works. That onus is on the designers of that software.