Often distributed systems are considered robust if one of the components halts. But a failure mode that is often neglected is when a component continues to operate, but incorrectly. This can happen due to malicious intentional compromise, or simple hardware faults, misconfiguration, or bugs. Unfortunately, there is no single add-on to designs that will fix this case. This talk presents three very different systems and how they each handle resilience despite malicious participants. The problems, and the solutions, are very different. The important message of this talk is that there is no one solution, and that this case must be considered in designs.