a) random fluctuation (most probable)
b) someone developed ASICs three months before BFL and mining for themselves
c) someone is attempting 51% attack to destroy bitcoin
Is (a) most probable? Perhaps one our local math whizs can figure the statsitical likelihood of a 50% move in < 24 hours. I don't think it's even remotely likely given a base of this size.
I'm leaning towards some other company with an ASIC design. They likely pushed really hard to get them active as the price was rising. Even still, we ought to see that skew the pool stats unless they intentionally spread them over all pools to remain hidden.