Note that little of this really helps the students that it is supposed to help, because as you wisely point out, raw HTML is almost by definition extremely accessible. I work in a field that uses Latex and the source code of Latex should also be considered more accessible than the compiled pdf. But for university administrators the only important thing is that the accessibility metric that appears (or used to appear, before today!) on Canvas shows 100% accessible.
Nobody has infinite energy, and disabled people don't have infinite social capital. It's a shame when energy from that shared pool gets spent on things that don't really impact meeting people's access needs.
And the other thing is that everyone's access needs are different. It can certainly be useful to try to set a baseline or propagate common guidance. But the most important thing, especially in a university setting, is for instructors to be flexible and responsive and for classes (and non-teaching workloads) to be structured in a way (e.g., small enough) that supports that.
I think metrics like "100% accessible" might even be dangerous. It makes it easy for able-bodied people who aren't in direct contact with disabled stakeholders to pat themselves on the back without actually knowing what's going on.
Bleh. Good luck doing right by your disabled students and disabled colleagues, and good luck resisting the bullshit.
That said there is certainly a lot more work that needs to be done in this area. Hopefully these regulations over time bring out practical positive change. Time will tell.