"Some of the best additions were API extensions, but have not been capitalised upon because developers wont support a subsection of the market based on a diminishing number of old machines."
ROL were in a bit of a lose-lose situation, so you can't entirely blame them for trying. If they add application-style features to the OS, they'll get burned by commercial developers. If they add API and lower level functionality, they get burned by customers when no developers use the new APIs. If they had the resources to make the system PMT+fully memory protected, then you'd probably lose at least 75% of your software through compatibility breakage.
The process of 32bitting software turned out to be pretty trivial, which did Castle a lot of favours, but if you think about it, not a lot of the new APIs in RISC OS 5 are used outside of the handful of new hardware drivers - and this is probably for the same reason developers are unwilling to commit to using the Select-only APIs. The Iyonix is faster than the RiscPC, but clearly not fast enough for the mjaority of users to part with their cash for the price tag it carries.
AdvantageSix have been pretty crafty in aiming their machine at a middle space in the market with performance just below the Iyonix and above the RiscPC, and with a lower price tag than the Iyonix. And if you want things like an internal DVD drive, then you're welcome to buy an Iyonix. Hopefully that'll encourage people to upgrade to recent versions of ROS 4 and persuading developers into leveraging from it.
I can see this turning into a survival of the fittest, where the OS with the most users will continue, which is a bit unfortunate as it leaves potential buyers hoping they jump onto the ship that doesn't sink.