I'd assume that this is a direct consequence of the impact mutations can have during short spans of generations. The closer you are to a local optimum, the more mutations you need to get into different (albeit better) optima.

Essentially, the step size of the optimisation process is usually too small to make this jump, you need a lot of luck to make it work (since any transitional generations have to stay alive long enough to reproduce and outcompete/find a new niche) - which automatically gives the rest of the ecosystem time to "catch up", changing the landscape of the fitness function and thus providing new pathways to better optima.