Per the article, the vast majority of traffic is scrapers that are generating effectively random, permuted URLs. Even if they're already caching their popular-with-real-humans most recent data, that efficiency gain is a tiny drop in a very large bucket.