If there is a failover while the dbhash check is running, the shell_options will become stale so the donor/recipient connection strings may not correspond to the primaries.

Furthermore, if a primary node is killed and takes a while to start back up, the hook will also continuously fail to connect to it.