The use of redundant systems has become increasingly common in AI engineering to improve reliability, but the extent of their capabilities have yet to be fully explored and tested.