The SRV records result processing code tries to filter out duplicate entries, but fails to do so properly. The biggest bug is that the would filter an entry as duplicate even if just the port numbers match:

if (server->port == reply->port) {
    ret = EEXIST;
    break;
}

This might create false-positive DEBUG messages for cases where SRV query expands to the same entries that are hardcoded in the config file as fallback hostnames.

Instead of fixing the duplicate detection, I'd like to suggest we remove it altogether. The hardcoded server hostnames are only meant to be fallback for cases where SRV records fail (a roaming laptop)

Consider this example: a config file contains "ldap_uri=_srv_, foo.example.com". SRV query expands to "foo.example.com bar.example.com". With our code, foo.example.com would be filtered out as a duplicate and the resulting server list would contain "bar.example.com" as the first entry, which is wrong.

The downside would be longer timeout as one server might be contacted multiple times.