When OS X came out (both the Mac OS X Server 1.0 and Mac OS X 10.0 varieties), the UI might have looked different, but you were basically running OPENSTEP 5 with a new coat of paint. Haiku could do a new theme fairly easily at this point, and there are already quite a few touch-ups of the new-slab-of-paint variety (e.g., subpixel antialiasing, support for tiled windows, various improvements to OpenTracker).
The other thing that OS X brought to the table were new APIs, but most of these APIs weren't really user-visible per se. (E.g., QuickTime was a genuinely new addition to OPENSTEP, but the end-user would just see it as more videos played now, and they played with lower CPU usage.) Here, too, Haiku has delivered: it has new APIs for component layout, new APIs for end-user notifications, and more.
So your comparison to what Apple did with OS X is actually quite apropos. Be just hasn't felt a need to overhaul the UI as radically as Apple did in going from OPENSTEP to OS X--in large part because, while some of those changes were functional, many of them were more about making a statement than being genuinely easier-to-use. (Note how much OS X 10.6 and 10.7 have largely reverted to a kind of brushed-up version of the old Platinum interface: no more pinstripes, flat grey buttons in most cases, flat grey title bars, etc.) So I think they're doing exactly what they ought to be doing.
It's entirely time and money. They only have so many devs, most of whom have day jobs, and pretty limited funding to support them, so the UI (and many other polish-and-details things) had to take a back seat to more important core functionality.
Because it is extremely much work, both in the research/prototyping/design and in the implementation. And when you are done, there are no apps that supports it.
I assume they don't exacly have a huge community of active devs. combine that with the fact that user interface design is hard and needs experienced people working on it and you got the reason for cloning a perfectly functional UI instead of completely messing up while innovating a revolutionary new one.
