One thing Swift will address... There are currently 3 memory management models in use in Objective-C, and for some of those models, you don't get a retain count automatically (for example, this is the case for a number of collection objects when doing an insertion).
Swift has the opportunity to rationalize this, which is not something you could do with the Objective-C libraries themselves, since doing so would change historical APIs and thus break old code.
It wasn't really until Metrowerks basically became incompatible with the Intel switchover and the 64 bit support had to drop certain types of support from Finder due to 64 bit inode numbers, and while I happily would have made them new header files so that they would have continued to work with the UNIX Conformance work, where Ed Moy and I basically broke their local private copies of their header files, since Motorola sold off the Intel version of the Metrowerks C the week because Apple announced Intel, it was pretty much DOA at that point.
So it basically took an Act Of God to get some people to get the hell off some of the old APIs we had been dooming and glooming about for half a decade.
Swift is another opportunity for that type of intentional non-exposure obsolescence to clean up the crappy parts of the APIs and language bindings that haven't been cleaned up previously due to people hanging onto them with their cold, dead hands. Hopefully, they will advantage themselves of this opportunity.