You may know how to ride a bike. Can you reverse engineer and explain in an algorithmic way how to ride a bike? No. Does that invalidate the fact that you can indeed ride a bike? Of course not.
Edit: do not fixate on the bike. It was just an example. Write a symphony, create a joke, understand irony.
Dowsing for corpses, on the other hand, absolutely requires new physics, because the pseudo-physical explanations offered around piezoelectricity and people having different voltages are deeply inconsistent with our understanding of physics.
But riding a bike seems very easy algorithm. All these self balancing scooters, remote controlled toys, seems like perfect examples of what you get when you reverse engineer bikes.
Anyway, the fact that we can not formally prove all true things doesn't really matter. That doesn't make all things equally valid. We find supporting evidence, we build conceptual models, we test the periphery of systems.
You could observe the sun rising and setting long before understanding the revolutions of our planet.
If someone can do something, you can reproduce it. That's true for bike-riding, but apparently not true for dowsing.