'Business programming in a large' is orthogonal to the mathematic structures. I.e., in the naive view, deep mathematical abstractions should lead to beautiful small easy-maintainable programs. In practice, it's not: software systems in condensed abstracted style are often hard to support because of strong coupling. Math is like a crystal -- it has a stable structure that formed around the main idea. Business is like a natural object, a river or tree -- its structure formed by different thoughts, which often contradict each other.
We can think about metrics here -- the rate changes should be commensurate with the cost of ongoing refactoring.