Firstly, we can assume that the RiscPC-class machines are essentially one and the same, beyond a few minor tweaks for the CPU. That means ARM 6, ARM710, StrongARM, ARM7500, VRPC, and RPCemu are all the same thing. There has been no version of RISC OS 6 released for ARM9 machines, and only an unstable beta for RISC OS 4. So that's one machine, perhaps one and a half if you include the A9. RISC OS 5, however, runs on the Iyonix (with graphics card support, like you claim for the VPod), RiscPC-class hardware (half a point), the BeagleBoard (half a point), and all of the different hardware Pace used it on (we'll call this zero points, but pointing to its working). It'd also most likely work under VRPC too if it hadn't been intentionally nobbled to only accept ROL ROM images. That's two verses ROL's 1.5, and that's being mean to the RO5 side.
RO5 *already* supports different video cards. It *already* runs on more architectures. And you've completely ignored tlsa's point that nothing you link to proves, or even agrees with anything you're saying at all.
You seem to be fundamentally confused between the CPU and the system, as well as what a HAL is. The RiscPC is one system, which a choice of (very similar) CPUs. And I don't doubt that a graphics driver for the VPod couldn't be written for RO5 pretty trivially; the SM501's quite easy to drive simplistically.