Recount > Count?

Here's the thing I have always wondered about recounts: why are they presumed to be better than the original count? If you count one time and get result A, then count again and get result B, how do you know which to believe? Is there a third tie-breaking count? (What if the tie-breaker produces a third result instead of validating either of the prior counts?) Maybe the recount is done in a way to make it more rigorous than the first count, thus giving it more credibility?

This is not about the current recounts in PA, MI, and WI, but about recounts in general. I am not criticizing those recounts or the concept of recounts. I just don't have the answers to the above questions.