If the issue is health & safety, and the dangerous failure mode is pretty rare - then how much time should be spent in validation, testing, etc. - to make sure the fix you rush out the door ASAP does not actually make things worse?