It seems to me that RISC OS 6 should not be contemplated unless and until the separate developments of RISC OS 4 and RISC OS 5 come together into a single, unified tree (however feasible that is). If there is one new OS that can work on any modern hardware (Iyonix, A9, Virtual RPC and whatever else might emerge), and we don't have different people developing the OS in multiple opposing directions, then that's the time to call it RISC OS 6 (and to rejoice at the new-found unity it brings); not before.
Even if it were to happen, though, developers would still be faced with support issues for existing RISC OS 4 (whatever flavour) and 5 users. It would take a long time for there to be widespread adoption of any hypothetical RISC OS 6, and existing software would still have to cater for the machine it's running on, so introducing RISC OS 6 would add just one more combination to cope with, regardless of how good a move it might be for the future.
Anyway, practically, I would imagine that there would be some significant technical differences between Iyonix Select and RISC OS 4.4 as seen on the A9, even if the APIs are compatible and the actual 'bullet-point features' turn out to be much the same.
(Disclaimer: I'm speaking entirely for myself, not on behalf of RISCOS Ltd, Castle or anyone else. I can't comment on what they're doing; if I could, I wouldn't be posting this! )