One of the side discussions at last night's rocket society meeting had to do with climate change - and the character of the computer models that are being used to make those scary predictions for the next century.
It is of course well known that computer climate models still have pesky known bugs, the most conspicuous manifesting themselves as butterfly effects.
One member claimed that there were rounding errors involved, and that rounding one way gives runaway warming, while rounding the other way gives - no, not money in my bank account - runaway cooling.
Anyway, I had a sudden inspiration. What if the butterfly effect is real?
If it is, the whole problem can be solved with a handwave. You just need a good enough model to calculate just where and when to wave your hand....
(On a non-silly note, I've had enough experience working with floating-point numbers to realize that loss of precision can lead to dramatic effects in iterative calculations, and that order of evaluation can matter in practice, despite what you learned in math class. Is there really any truth to the rumor that the climate-modeling crowd isn't minding the quality of the numbers inside their models?)