It prioritizes the platform vendor's feature desires over those of the developer or the users' direct vote, but I think it's wonderful for the industry to have at least one vendor (and one who arguably has good taste!) doing so.
Nobody wanted a mouse until they were shown what you could do with a mouse and Aldus Pagemaker. Nobody wanted "undo" consistently implemented across all applications. And absolutely nobody wanted their applications to run in fully little "windows" on the screen.
From this cherry-picked example, I conclude that a good platform does allow the vendor to push certain features forward, but it also gives developers a certain amount of leeway to innovate.
This is why I'm happy to see both models exist and thrive. Android would look like an old Blackberry if not for Apple. And likewise Apple can pick up any advantage that Android has (other than a few, such as meeting the needs of markets outside Apple's target).
Resume is more involved, however: the application is responsible for (de)serializing its state. Again, for a document-based application that doesn't stray far from the "recommended" path, it should be fairly simple to implement. It's just a matter of the document and UI being auto-saved on termination, which requires little work from the developer. One very interesting side-effect of the Resume feature is that it removes the need for applications to stay running: the system can silently terminate unused applications, and bring them back if they're needed again. What's really exciting about this, is that it's turning applications into something that is "always there".