Just some random thoughts on this subject.
First comment. OS X on x86. It would not surprise me in the slightest to find out that Apple is actively maintaining a port of OS X to the x86 architecture. Why? Because there are certain classes of bugs that will show up when you port to different architectures which won't appear if you code on the one arch. Endian issues, for example, can be good to get rid of; it also helps to make the code that much cleaner (often, cute hacks which work on one platform won't work on another.)
Having said that, though, I'd be extremely surprised if Apple ever released it for public consumption. Darwin, yes -- Darwin is out and available for x86, but the Aqua layer is not, and probably never will be.
Second comment. Application support. NeXTStep has, right from the word go, supported so-called "fat binaries": binaries with support compiled in for multiple architectures. This is still present in OS X, and is likely to be put to good use once apps start fully supporting the G5, so that G4 support doesn't have to be completely thrown out the window. The same setup can be used for x86 support, but for that to happen, Apple would have to release x86 cross compilation suites to application vendors. They wouldn't release OS X on x86 until those suites had been seeded for some time, to allow the application marketplace to get ready; I can't see anything of the sort happening any time in the foreseeable future.
Third comment. Apple's likely future paths would almost certainly include Cell (as other posters have said, the core CPU in Cell is a PowerPC derivative; the ancillary CPUs aren't, and would be used for other purposes), as well as whatever IBM will call the cut-down POWER5 (the 980?) and later CPUs. It's a question of what they can engineer into the case without compromising the stability and reliability of the system as a whole. Cell is actually an interesting one; it would likely be used in a "G6" line, because of the ancillary CPUs, unless IBM or others release a cut-down Cell. Too early to tell yet.
Finally: licensees of OS X on other PPC-based systems. I could see Apple doing this down the road; just not yet. Apple's marketshare is too fragile right now. If it jumped to 10%, I'd not be surprised to see a limited clone market be permitted, with appropriate QA advice from Apple to try to maintain the stability of the platform, in an attempt to broaden the appeal of the market; but it's not something Apple would let slip until it actually happened. Having said that, it wouldn't surprise me if this didn't happen either. Long term (five to ten years), I'd call it a fifty fifty shot; I wouldn't be placing any bets at this stage, either way.
