We require software to be resilient because it is used as a building blocks for large (sometimes exteremely so) systems. The deeper in the stack something is, the more costly the failure. A human may not input any of those strings. Another software (which does not reason in any way about the data it is processing, unlike the human) may do whatever.