Part of the reason Castle decided to open RISC OS up is to make it more attractive to potential users. RISC OS is already used in STB's but the advantages being eroded as alternatives get these features e.g. fonts that look good on TV and being first into the market.
Designing STB hardware with dedicated MPEG video decoders (or other custom hardware) would require significant customisation of RISC OS. Providing a license that allows third parties to experiment with the OS means they can do so in private and with a choice of developers (not just Castle) to work on the project.
Linux is definitely competition to RISC OS and has compelling features. To make RISC OS a more viable alternative then increasing the speed and reducing the cost involved in migrating to other ARM processors is critical. At the same time RISC OS must offer features that differentiate it from other OS's targetting ARM.
The STB market is huge think satellite decoders, TV on demand in hotels, Inflight entertainment systems, TV's with built in digital decoders, HDD recorders etc.