In one of the classes on Lebesgue Integration that I was attending , there was a student / executive who was rather pained by the academic side to probability analysis. Professor was trying to prove that a vector of n dimensional normal random variables is also a normal variable if every linear combination of the n components is a normal random variable.
At some point of proving the statement, the student felt frustrated and blurted out that he wanted more practical applications than theory. At that juncture, one could easily see that professor was rather disappointed with the perspective of certain students. She felt that any learning should make a person comfortable with the theoretical underpinning as well as the practical aspect of it. One superb point that she made was, all the models, all the applications in the world would change , but it is very unlikely that the theory on which the models are based is going to undergo rapid transformation. Innovation comes from understanding theory behind the models.