
Ping times were perfect and bandwidth looked healthy, yet users reported that every page took forever to load. The problem was not packet loss; it was DNS. A resolver in one region had a stale cache and retried lookups for an internal service, adding hundreds of milliseconds before the browser even sent a request. We confirmed it by timing name resolution separately from connection setup and TLS handshake. The fix was small: refresh resolver caches, shorten TTLs for critical names, and monitor DNS latency as its own metric. This case is a reminder that network health has layers. A green link can still hide slow name resolution, certificate negotiation, or connection retries. When you investigate a slow path, which layer do you check first, and how do you separate it from the rest?