Snow leopard was, as you said, necessary in anticipation of the architecture change.
Now there's no such change, but instead AI, this weird new cross-cutting but fuzzy function touching everything that no one has ever used reliably at the scale of Apple devices. AI is impossible to reliably test, and all-too-easy to get embarrassing results. I'm glad Apple recently tamped expectations.
The relatively loose concurrency model in Apple's ARM has made it rival the network in introducing new failure modes Many quality issues cited have their root causes in those two sources of indeterminacy.
Amplifying these are the organizational boundaries driving software flaws. Siri as a separate organization with its own network-dependent stack is just not viable for scattering AI. Boosting revenue with iCloud services makes all roads run through the servers in Rome, amplifying network and backend reliability issues. I also suspect outsourcing quality and the maintenance of legacy software has reduced the internal quality signal and cemented technical boundaries, as the delegates protect their work streams and play quality theater. The yearly on-schedule cadence makes things worse because they can always play for time and wait for the next train.
And frankly (to borrow a concept from Java land), Apple might be reaching peak complexity. With hundreds of apps sporting tens of settings, there is simply no way to have a fast-path to the few things different people need. Deep linking is a workaround, but it's up to the app or user to figure that out. (And it makes me livid: I can't count how many important calls I've missed by failing to turn off "Silence unknown callers", with the Phone app settings buried 3 layers deep ON MY PHONE)
A short-term solution I think is not a rewrite but concierge UI setup: come to the store, tell the "geniuses" exactly what you need, and make shortcuts + myUI or whatever is necessary to enable them to make it happen. Then automate that process with AI.
That's something they can deliver continuously. Their geniuses can drive feature-development, and it can be rolled out to stores weekly and -- heavens! -- rolled back just as quickly. Customers and employees get the excitement of seeing their feature in action.
The model of sensitive form-factor designers working in quiet respectful collaboration to produce new curves every year is just wrong for today's needs. All those people standing around at Apple stores should instead be spending an hour or more with each existing customer designing new features, and they should be rewarded for features that take, and especially for features that AI can incorporate.
On the development side, any one should be able to contribute to any new feature, and be rewarded for it. At least for this work, there would be no more silos, and no massive work streams creating moral hazards.
The goal is to make software and a software development process that scales and adapts. It may start at 5% of new UI features, but I hope it infects and challenges the entire organization and stack.
Granted, it will take a famously hub organization and turn it into a web of hubs, but that in itself may be necessary for Apple to build the next generation of managers.
Look for how today's challenges can help you build tomorrow's organizations.