I don't see the POINT of the ARM only restriction. As most of the source is ARM asm it would be an enormous task to port the entire OS to x86. Remember, although someone might discover that the released source is only 40% asm (that's a completely made up statistic) what would need to be ported is the core of the OS itself. Many of the ancillary support apps such as !paint might contain significant amounts of portable C code but I suspect that more than 50% of the OS itself will turn out to be asm.
It would probably be just as easy to recreate a RISCOS-like OS from scratch (perhaps making use of some freely available components), and re implimenting the API. I think that this would would be about same amount of work and produce a better result.
Unfortunately, the announcement that Castle won't ALLOW people to port to different platforms has turned a lot of people against the project from the start. And for no real gain from Castle's point of view.
At least this new development offers a new way forward for the OS.