We consider a model of governors serving a sovereign prince, who wants to deter them from corruption and rebellion. Governors must be penalized when they cause observable crises, but a governor's expected benefits must never go below the rebellion payoff, which itself is better than what any candidate could pay for the office. Governors can trust the prince's promises only up to a given credit bound. In the optimal incentive plan, compensation is deferred until the governor's credit reaches this bound. Each crisis reduces credit by a fixed penalty. When a governor's credit is less than one penalty from the rebellion payoff, the governor must be called to court for a trial in which the probability of dismissal is less than 1. Other governors must monitor the trial because the prince would prefer to dismiss and resell the office. A high credit bound benefits the prince ex ante, but in the long run it generates entrenched governors with large claims on the state. Low credit bounds can cause the prince to apply soft budget constraints, forgiving losses and tolerating corruption for low-credit governors.