Of course, in theory, A/B as the name implies is about testing two different approaches, so again, in theory, any stuff completely different can be seen as an experiment for A/B. The problem is that humans have been trying different ways to do things for a loooooooooong time, and the whole point of trying to teach and learn A/B is to test things, measure, and walk towards your goal.
So yes, you can "A/B test" two completely different things, but it will be so. much. harder to actually know what it worked. For example, you're A/B testing a bicycle and a Ferrari Testarossa. D'uh... obviously they'll work differently, but what caused? Price, color, speed, brand?
That's why when A/B testing is mentioned it's about small, isolable, repeatable, incremental and evolutionary approach. Because it's a technique with this purpose in mind.
I guess what I'm asking is this: Is A/B testing an evolutionary approach to developing a product? Are "revolutionary" leaps possible if you strictly follow the A/B testing methodology? (My guess: probably not, which implies that most people don't follow A/B exclusively, which sounds like common sense).