Most of my work in military R&D was dealing with the unknown unknowns. The "unk unks." These are the ones that will bite you when you don't expect it. The first goal is to convert them into known unknowns. "I now know this phenomenon exists. I don't know the limits of this phenomenon, but I can design around it." After that you want to convert them into knowns. "I now know the range of this phenomenon, and I can exploit it."
Failure to recognize that there will be unk unks is a recipe for disaster.