Behe actually has come up with an answer to this. It’s a silly one, but an answer nevertheless. Behe essentially argues in Darwin’s Black Box that simple irreducibly complex functions can evolve, by which he means systems with 2-3 components, but that as the number of components of the system increases, he argues, the probability of evolving it goes down.
It is obvious what the issue with this kind of argument is. Essentially Behe has sort of argued that his concept only really takes effect once the complexity of the system is beyond what we should normally expect to see happening in human lifetimes.
Sure, we can see a 2-3 component system evolve in some experiment, but what about a 20, or 50 component system.
With this very arbitrary limitation, Behe has made sure to put the goalposts beyond the results of what could be reasonably expected to be observed in an experiment happening in a couple of decades of study. In this way Behe will always be able to argue that more complex systems have never been observed evolving, and that only “modest” and “simple” results are achieved in lab experiments.
It is difficult to resist suspicion that he did this intentionally just so he could fall back on the age-old creationist “observational” vs “historical” science objection.