@jonny What makes that difficult for me is that it's often really difficult to reason about the consequences of wrongness in a system, and whether it might affect something that really *should not* be wrong (eg. security or safety systems) in an unexpected way