I'd suggest unifying the OS, then looking towards emulation. Hardly anyone wants to pay huge amounts of cash for RISC OS - it's so behind in terms of modern operating systems and desktop environments that it's embarrassing.
Oh and sort out the crappy cooperative multitasking which makes the whole thing feel so much worse. Locking up the system for 10+ seconds to unzip a 5Mb archive really isn't acceptable these days in a desktop environment. I actually prefer using the command line to the desktop simply because I expect that to single-task and don't mind the wait - then again, I prefer to avoid using RISC OS for anything remotely significant anyway. It's simply a morbid curiosity thing, wanting to know how much the OS has progressed in 15 years and quite frankly it hasn't made much.
Basically I'd say that anyone wanting RISC OS on it's own PC hardware is as delusional as the idiots at Castle. Asking people to pay through the nose to pay for it is also just as laughable.
I suggest that anyone who's serious about developing for ARM based machines should just abandon RISC OS - it's a waste of your personal time and money. You're not martyrs and holding the candle for a system which no-one really gives a s*** about will only get you burned as so many have already learned - and there's plenty of money to be made from programming ARM p;processors elsewhere.