I think most people, when requesting a citation in this context, would also be fine with a description of an experiment (preferably a repeatable one) and a record of the empirical result of that experiment. That's the "base case" you're talking about in the otherwise infinite regression of citations.