I don't see the argument of "it's never going to be good enough to be the mainstream desktop" as reason enough to warrant abandoning this project. That's a bit of a short-sighted view when you take the entire entertainment market into account.
To start with, presumably JRiver's refactoring of the MC codebase made it economically feasible to explore new options, even if there's a bit of risk involved. We have OS X to thank for that, I suppose
You also need to remember that Linux isn't just the "desktop Linux" of years past. There is a huge range of devices out there using Linux as the OS, from routers to media centre units to, as recently announced, future home consoles (the upcoming range of Steamboxes). I can definitely see a market for people buying MC for these devices, or JRiver selling custom hardware of their own (which would be a lot more flexible and cheaper than shipping barebones PCs running a heavily customised Windows install).
Also, you need to remember that having MC on more OSes is *good* for the Windows version. This might sound counter-intuitive, but the more platforms a piece of software is on, the better the codebase usually becomes, since developers tend to try to rely on fewer and fewer platform-specific hacks. We've seen time and time again since JRiver started their massive porting effort that this leads to things like better performance across all OSes.
Really, Linux opens up a huge range of opportunities, and I'm excited to find out where we'll see MC running a year or two from now
It'll be in my Linux media centre come October, that's for sure