Jon Blow said--I believe--in a tweet that he decided against academy when reading grad students' papers proving the optimality of some algorithm where, in practice, caching was of prime importance (or something to that effect).