if one instance dies, the extra pods may ensure that there is no actual outage while the failed pod restarts