I'm afraid I see no future for bespoke ARM based hardware. Back when a decent PC cost £1000 and a RISC OS machine cost £1500, you might be able to justify it. Now a new limited production run RISC OS machine is still likely to cost at least £1500, but a reasonable x86 system can be bought for £300 including monitor, which has a vast performance advantage over any ARM system. Put an emulator on it and it will run RISC OS at an equivalent or better speed.
Wishing to retain the native hardware is a purely emotional attachment, ARM aren't in the business of making desktop class processors, so there aren't any economic or technical reasons to build a machine around them. Even the newer dual core processors won't give that much of a performance boost, as its one thing to have two cores, quite another to modify the arcane internals of RISC OS so it makes any significant use of them. Indeed a few tweaks to VRPC would give a far greater boost in performance on multiple cores.
Then there is the question of what a RISC OS machine is used for, which is to run legacy RISC OS software, with the many gaps in the portfolio, such as a full web browsing experience, having to be performed on another platform. There is little chance of new hardware spurning a significant increase software development, which now consists of the sterling efforts by few remaining developers producing upgrades to existing applications, and a few new utilities produced by hobbyists. Justifying a new machine on this basis when you already have other machines is impossible.
I hope that when ROOL have release all the sources to RISC OS 5, people will port it to some interesting ARM based CE (consumer electronics) devices. But it isn't likely to be a commercial operation, by the time its done the item would be passing out of production, so it may be of benefit to a handful of people. I'd also like to see Virtual Machine standard of emulation available for all platforms, including Linux with Aaron is so reluctant to support. There is much scope for improvement over Virtual RiscPC which has remained largely unchanged in 5 years due to lack of competition. RPCemu with a good JIT running on Linux, would be a much better prospect to allow users to retain access to the classic RISC OS applications they are familiar with, but also enjoy all the benefits of a vibrant rapidly developing platform, free from Microsoft hegemony.