I believe that Radiola is correct that fields was intended as an interface to pseudo-hashes. I remember documentation urging the use of fields because it was specifically intended as a forwards-compatible interface that would still work after pseudo-hashes were removed.

Pseudo-hashes seem to have been recognized as a bad idea almost as soon as they were implemented.

You are also correct that fields has nothing to do with object methods; Radiola is confused on this point, although the documentation is also confusing here, since fields is documented in OO terms.