In fact, these are incorrect log entries, these error should be "could-not-connect"
I use sort of reverse-engineering to build mapping of pycurl error codes to their short text abbreviations with the code:

for key in dir(pycurl):
    if key.startswith('E_'):
        abbr = key[2:].lower().replace('_', '-')
        ERROR_ABBR[getattr(pycurl, key)] = abbr

The problem is multiple E_* constants points to same numeric code, so E_COULDNT_CONNECT is overwritten with E_MULTI_ADDED_ALREADY