Then there is the general lack of configuration tools for programmable game controllers.
I'm sure Valve (and/or future console manufacturers) will work with the necessary upstreams (X, kernel, etc.) to get better drivers and configuration tools if they need them...
Extremely exciting news! This will surely change the Linux Gaming (and Linux Desktop) landscape.
I mean, sure, there are actually more games on OS X now. Steam had a positive impact. OS X is still nowhere close to dethroning Windows as the PC gaming platform of choice, but OS X users have a lot more options to have fun these days than they used to. Linux will likewise see a positive impact from Steam, but it will continue to be missing large swaths of the best games available, just like OS X.
A Steam Machine (please let that be the actual name) would be another boost. Typically new consoles have flopped, independent of technical qualifications but we are seeing a recent strong shift away from the traditional console market, so there may well be a huge opening for something like the Valve console (or Ouja). It's all going to come down to the actual game catalog on the device. If you can't get Activision and EA and Ubisoft and so on 100% on board with a Steam console (including Linux ports), it is not going to do well. Valve will need to give those companies incentives, which even on the PC space they're having a huge problem doing right now (notice how many of the most popular PC games are not available on Steam these days, including many that have console ports).
Unfortunately, a large part of that is going to be how tightly they can lock down the machine for DRM reasons. Game publishers put up with ridiculous rates of piracy on the PC because they don't have much choice (DRM on a PC just doesn't work to stop anyone but very casual pirates). Any new platform, however, is just a new risk to them; if they can't get things locked down, they have no reason to spend money to port to a platform just to see 90%+ of their efforts stolen. Especially when they can keep making lots of money with the new Wii U, the next XBox, and whatever Sony is up to and get much lower piracy rates (it requires more effort to even use pirated games on those platforms, effectively barring the large masses of casual pirates).
And since Valve's console's success will be tired _entirely_ to the game catalog on it, getting those publishers on board is the #1 most important thing for them to do as a for-profit corporate entity launching a new for-profit gamer-oriented consumer electronics device. They are going to lock down their console as much as is required to satisfy those publishers, or they are going to fail miserably. Maybe they can convince those publishers that they don't need to lock down the console much, and that's great. Quite possibly they will not be able to do that, and the new Valve console will only be good for a handful of titles compared to its competition.
That doesn't even include exclusive titles to the other platforms, such as Nintendo's beloved catalog, Microsoft's impressive catalog of titles from MGS and subsidiaries, or Sony's critically-acclaimed first-party titles. PC does have exclusives, but most of those _aren't on Steam_ (even most of Valve's own titles have XBox ports), or are smaller indie games that will hardly push sales of a new hardware platform (and if they do, Valve will be competing with Ouja, not traditional proprietary consoles).
tl;dr version: Valve has to suck publishers' $%#@ for a new console to succeed, as success in the consumer device market will have nothing to do with openness, Linux, or Valve being Valve.
Take the Linux xbox driver, for example. The last time I looked at it, the driver doesn't support audio/headsets at all. It doesn't support battery lifetime notification at all. It has no notion of which controller is 1st player, 2nd player, etc.*, so each individual app has to redefine which is which. The driver does support the LEDs on the device, but it doesn't enforce any player-oriented friendly use of those lights, but rather just lets apps do whatever random confusing inconsistent useless crap it wants to with them. Well, sort of, since the LED device and the input device are completely separate devices in the device tree, and actually correlating them together is non-trivial, requires deep Linux kernel knowledge and some voodoo, and isn't wrapped by any existing developer-oriented library.
Many of the other gamepads are just supported by the generic joystick driver and hence are also missing all the advanced features and proper system-integration necessary for a quality gamepad gaming experience, as supported by the consoles and Windows.
* Some people don't believe me about this one because the Linux kernel driver will set the player-indicator lights on controllers when plugged in, so if you plug in two controllers, one is lit up as player 1 and the second as player 2. The problem is two-fold: first, there's no way to read the status of the lights nor does the Linux driver expose any property as to which number it assigned the controller, so apps have no way to know which controller is which. Second, Linux doesn't actually assign a player number, but rather fakes it with a cheap hack. The driver just does a "indicator_light = (global_controller_counter++) % 4" when a controller is plugged in. You can see this by unplugging a controller and plugging it back in (it'll go to the next player indicator, rather than remain the same number); more damning, plug in two controllers, then unplug and replug just one of them several times, and you'll quickly get both controllers to have the same player indicator lit up.
I'd like to see steam with some level of xbmc integration, all i'll need then is a wireless nostromo so i can sit on the lazy boy in front of my projector munching n00bs