Most graphics cards put out fill what would be called the "bleeding edge;" that is to say, they are designed to be close to BARELY ready for full production. Over time, tweaks to design and production methods will improve effectiveness and efficiency, though in the fast-paced market of graphics cards, this rarely comes in time to have hope of competing with what's new; the best that gets seen are some quick die shrinks (usually half-node ones) that often also serve to carry "experiments" in design for future cards.
If Galaxy manages to make this a card that goes out smoothly, it should actually sell well. That is to say, if they get, say, better performance than a pair of 9800 GTX+ cards in SLi, and get a low TDP (at least idle TDP) while they're at it, and can produce and sell them cheaply, it'll definitely have a market.
Remember, the performance crown-wearer never makes the most money; it's the more practical sub-top and mid-range cards that make up the most sales and money at retail. G92's a technology that's been tried and proven... And used again, and again, and again, even if to the dismay of many enthusiasts. So if it works, why not give it another shot? The only potential problem is the PCB length, which I'll grant has me a little concerned.