One approach that some users who want this mechanism use is the "on-error sudden-death" mechanism: the server responses are inspected, and upon error (from sele

One approach that some users who want this mechanism use is the "on-error sudden-death" mechanism: the server responses are inspected, and upon error (from selectable types), the server can be marked dead. If all of them are dead the backup server will come into the dance and the last attempt will end up on it.