Sod porting an existing version of RISC OS onto a machine which /already has one/.
Surely it would be more use to try to move it onto ARM-based hardware which doesn't currently run RISC OS? For instance, there are many ARM-based machines which run WinCE. These can run other OSs - there are several versions of Linux which run on them to some degree, e.g. Familiar, Debian and OpenZaurus. Ergo, there is freely-available information out there on how to access the h/w of these machines. If there's any commonality to that platform, a port to this would be more use. While RISC OS might not be much use on an iPaq, it would be very useful on subnotebook-class machines, such as the Compaq Aero. Alternatively, Simtec's ARM development boards spring to mind.
If something must be dropped to free up development effort, then I'd ask why maintain Norcroft C/C++? Might it be possible to move to GCC instead, thus getting the benefit of a lot of external development work? BeOS did this around the move from r4 to r5, and it is, for example, possible to compile a compatible, working Linux kernel using Intel's C compiler instead of GCC.