I think the fundamental barrier is that I have yet to see a system where mouse and touch can coexist as first-class input methods. Either your UI is optimized for touch with large input buttons and heavy reliance on gestures, or mouse with small input buttons that require precision to interact with and keyboard shortcuts for efficiency. The cognitive, not to mention physical burden of transitioning between a pointing device and a touchscreen means that users will favor one over the other. And if your UI has to target both audiences, then you're going to have to figure out how to seamlessly transition your UI or provide 2 parallel workflows, and at that point you might as well just segment your product.
It's easy to blame "Apple greedy" but optimizing either device to support an alternate input method degrades both. Apple is (supposed to be) all about a "polished" experience so this doesn't mesh with their design ethos. Any time I have seen a desktop environment get optimized for touch, people complain about it degrading the desktop experience. MacOS isn't even there yet and people are already complaining.
There are plenty of good AUs on the App Store (to name a few: DM10, Sonobus, the recent AudioKit modeled synths), but yes the selection of AUs on desktop is far greater. Most AU developers aren't going to pay the developer fee and go through the effort of developing, again, an entirely separate user interface, not to mention go through the app store approval process, to target a smaller market. It's a matter of familiarity. Just because your workflow depends on products that don't exist on iPad, doesn't mean that someone else's workflow isn't entirely productive without it. The entire industry is built on path dependence, so it's no wonder that software that has codebases that span decades and depend on backwards compatibility, i.e. the music production and CAD software, are not finding a lot of competition in the mobile space. Apple isn't designing their next unibody Mac chassis on the iPad, but that doesn't mean that a small business that makes 3D printed widgets isn't going to be happy using Onshape.
To be clear: I don't think an iPad is a _substitute_ for a desktop machine in most professional workflows. Partially due to path-dependence, and partially due to the greater information density that a desktop environment affords. But there are some workflows where the iPad feels like a much more _natural_ interface for the task at hand, and then that task can be transferred to the desktop machine for the parts where it isn't.