Is it not currently the case that there are redundancy tests (i.e. overconstraint tests / unit tests) on election results already? Would a manual miscount of a

Is it not currently the case that there are redundancy tests (i.e. overconstraint tests / unit tests) on election results already? Would a manual miscount of a vote have a near-nil chance of getting caught? If so, I totally agree we need to have samples of all elections verified to at least get bounds on error rates.