One of the other teams submitting a SIGGRAPH paper this season is in serious trouble. They've had to back off their claims and are scrambling to assemble decent results because of some assumptions they made. I won't go into details until after we get the reviews back, but the basics are that they expect nice, clean input data for their program. They're using the power plant model as one of their major tests. The power plant is anything but clean and well-behaved: it has every single pathological case I've ever heard of, and doubtless a few that I haven't. I said in the acknowledgements for my dissertation, "Thanks to the anonymous donors of the power plant model, which for five years now has been the standard against which algorithms have been tested and, occasionally, broken." That's what's happening to them right now. That's what (sort of) happened to me this time last year.
I think I'm glad I didn't try to add any major new bells or whistles this year.