With all the discussion raging in the blogosphere (see, e.g., slashdot, coturnix, evolution denialism, the futile circle, Nature, etc.), I think I should take one post to emphasize the scientific aspect of our work and another to emphasize the speculations one might start about our work.
This is the scientific post.

Scientifically, the most important aspect (which understandably got a little buried by the media) is that we found evidence for a brain function which appears evolutionarily designed to always spontaneously vary ongoing behavior. There are several lines of circumstantial evidence to suggest that such a function may be very widespread in the animal kingdom, including humans.
We found this by mathematical analysis of an old experiment. My thesis advisor (during my graduate studies in Würzburg), Martin Heisenberg, devised the "sensory deprivation experiment" over 20 years ago. His postulation then was that what the animals do should best be classified as what he called "voluntariness". Back then, many of the computational tools did not exist that we used now to analyze the behavior further. Actually, the torque meter I use has an inventory tag from the "Max-Planck Institute for Biological Cybernetics" from 1963 (I got it on loan from Martin Heisenberg - many thanks!). So the method is tried and tested. All our mathematical analyses have been published previously in peer-reviewed journals as well, the nonlinear procedures twice in Nature and once in Science. So all the methodology is tried, tested and sound.
Fly brains are designed by evolution in a way that makes them spontaneously vary their behavior in a nonlinear fashion. And they do this not only without external stimuli, they actually also do this when they fly towards a stripe, or in a randomly dashed environment or when freely flying. It appears that many behaviors of the fly are superimposed with a kind of spontaneity the structure of which has been found in many animals, including humans.
The interesting questions are: Why would so many brains have this function? Wouldn't it make more sense to clamp down on variability in order to be able to behave exactly as required?
Fortunately, there is plenty of literature on why indeterminate, varying behavior is evolutionarily advantageous. However, only rarely has this literature postulated spontaneity as the root for this adaptive variability.
That's the gist of it, leaving out the many interesting tangents I could go into.