Learning without thought is labor lost, Confucius said.

At any rate, symbolic representations of knowledge should be committed to memory (or pad and pencil) at the time of understanding (Hebb's rule), or else your "memory" will be isomorphic to random symbols and therefore inaccessible to future thoughts.

That said, some of the greatest mathematicians (e.g., Gauss, von Neumann, Dieudonné) had phenomenal memory, so I suppose rote memorization isn't a terrible hack to try to approximate the advantages this kind of gift confers, but only in a pinch, since these mathematicians were keeping much more than definitions and statements of theorems in memory.