Overfitting is when your model is over-fit to the data you are training on [1], but what we're talking about here is slightly different -- the EA has maximized the given objective function correctly, but it turns out that your objective function is corruptible: A high value on your simple measure doesn't match the intuition motivating it.
You had something particular in mind when you set the objective function for your biped robot to be distance travelled before it falls. Finding an exploit in the underlying physics engine that allows the biped to explode and travel quite far -- while indeed excelling at the function, is far from what you had in mind. Yet it does reveal a vulnerability in your physics engine, which while unexpected, may be valuable in its own right, as the application the post's author argues.
This isn't entirely unique to EAs -- but it seems to happen more frequently in the kinds of domains EAs are often applied to -- complex simulations where the gradient of an objective function is difficult or impossible to calculate.
I think he had something else in mind, somewhat related to overfitting -- you're maximizing (too much) a metric over an unrealistic solution space, and that solution when applied to a real solution space won't behave as intended.
So those are "overfitting" for the exact quirks of the engine. I agree, it's a stretch, but not unlike in ML, it shows that optimizing too much will show quirks if your 'objective function/simulation' pair is not exactly equal to the 'desired objective function/reality', or if there is an innumerable ensemble of situations you might expect the solution to behave well.
It happens in EA because the researchers don't account for over fitting. The experimenters often do not validate so the population converges to the training set maxima too far. If you had 2 simulations of walking. You would select and mutate based on one sim and stop the evolution when performance degraded on the second sim. Not blindly maximize the one measure.
if validation were performed the population would not evolve into the cracks of unlikely physics loop holes.
