GuestX: "...is less a financial restriction and more one of control"
In my opinion, no, this is not the intention. If the software is under the GPL then the viral-like nature of that licence infects software developed bespoke for the device in question. NRE is often a major proportion of the cost when developing a new embedded device. Many companies do not like the idea of putting effort into a product only to have to release all code to it - it's akin to writing off the NRE cost. Moreover, upon releasing sources, any potentially cheaper manufacturer in the world (notably China) will find it easier to copy the hardware, tweak the firmware in source form and erode the market started by the original innovator.
Some companies do release products around GPL operating systems like Linux, but they often either have a heavy string of software patents protecting other aspects of their design in the US (c.f. Tivo) or near-software patents good enough to scrape through in the EU, the product may be not in the slightest bit innovative anyway (e.g. a router), or it may be that the company is very large and feels able to defend itself on (say) the grounds of copyright violation rather than just accept a copy of its hard work appearing in the marketplace. Some even just ignore the code release provisions of the licence and cross their fingers hoping nobody notices! To avoid endless circular debate here, though, I would point out that there are of course countless permutations of this and no "right" answer when describing usage cases.
The reason why a royalty is payable for commercial RISC OS [Open] users is partly because those users can keep their code or modifications to core RISC OS private. They pay for what amounts to exclusivity. It's about control in a way, as you said - but not Castle's control, as you indicated; I believe it's actually there to grant more control to the user of the code.