There are a lot of theorems and results in mathematics that are very easy to state but often require a lot of advanced machinery to prove. We could easily come up with many examples of this: Dirichlet theorem, prime number theorem, and many other results in number theory, existence and regularity results of some PDEs, and the classification of, for example, finite groups using difficult representation theory. However, some problems that are solved using modern methods turns out to have an elementary solutions as well.
When we try to see how a theory is developed from axioms and assumptions, we often need to be more careful when using some "advanced" results, since we can only use things we have already proven; in other words, we need to avoid circular arguments. But now, let's consider a different scenario: problem solving. When we try to solve a problem, like things similar to Dirichlet theorem, we are applying theories rather than reconstructing them from axioms. In this case, how useful is it to solve a problem using only limited amount of machinery or even with elementary methods?
The question "how useful" might be difficult to judge, so actually, any ideas regarding why we need/needn't learn elementary proofs are helpful.