Does anyone here know of any existing formulation of the "law" that any sufficiently large dataset will contain every conceivable edge case?