A failure of governance if your goal is to have the best possible OS, and you have one person in charge who would rather not ship something than ship detrimental features, but that's not really how companies today work.
If instead we look at all of this as a company that doesn't really care about the overall product that much, and wants a chance of growth, then it all makes sense: Every team/owner decides that they want to ship AI in whatever bit of the OS they control, as it's a chance for relevancy with minimal downside. Then their boss realizes that they don't want to say no to anyone, and in fact might have the winning lottery ticket if more AI features are tried under them, and then you end up with the kind of disaster you see.
This isn't Microsoft specific at all: I bet many of us have seen this elsewhere, and even in different cycles. Everything is turned into a website whether it needed it or not, and then rewritten into a single page, because it was going to be revolutionary. Five different blockchain teams inventing use cases, including one spending a hundred million a year trying to make NFT videogames happen, and every project failed. This is the current governance standard in a megacorp.
People will only bother about the unstable mess when the risk is balanced, and they have as much to lose for ending up with an unstable mess as they have to win for risking instability for a half baked feature. Because I bet that, just like everywhere else, some people get promotions and large amounts of stock compensation for shipping a product before it proves to be good, so one can even be lavishly rewarded for failing.
So from where I stand, all of this is just Microsoft showing that they are just like everyone else. Given how fast the world moves, the governance you describe is rarer every year. So rare that even though I share your instincts, I am not even sure what "right" might be.