Gulli, yes and no. Basically I'm suggesting we go with the "insanely stupid" option. But wait! I agree that having the OS preemptive, multithreaded etc is desirable. The problem is how you get there from here. There is a chicken and egg situation with noone having the money or being willing to take the risk. So my suggestion is basically to smooth the transition.
As I pointed out, only very few apps actually need a lot of power - but those that do *really* do. And actually, the overhead in the app probably wouldn't be all that big at all - at least not for embarassingly parallel problems (which I think they tend to be for many of the things we want).
I agree we'd need to end up with a legacy environment, where all the old apps could live. But many apps wouldn't make the transition - and honestly, many don't actually need to. People only upgrade if there is a clear benefit. There would be an instant power increase since the obvious drains of the OS could be cherry-picked and temporarily put completely in the fingers. And then there would be a rapid and easy payoff for apps which need raw power. RISC OS could then be rewritten over time, with a clear benefit with every upgrade.