Intuition of double descent: when you have the about the same number of parameters as data points, a fit curve looks like one of those shitty "polynomial fits" in excel, where away from the data points it over and under shoots the curve wildly. Lots more parameters and this calms down to give smooth interpolation between the curves.

I can imagine some tasks where people have the same problem. You're overfit to a very specific task, but outside it "when you have a hammer everything looks like a nail" and you end up doing something dumb.