Edge caching and bot classification
A blanket TTL treated humans, search crawlers, and abusive bots the same — either over-caching stale inventory or under-caching and burning origin. We classified traffic at the edge and gave each class its own cache and block policy by page type. The hard part was false positives on legitimate crawlers; the tradeoff was operational complexity at Akamai in exchange for origin headroom and cleaner crawl budgets.