1. Data returned is very poor, basic and difficult to parse
2. To lookup you have to create a freedb id based on the track lengths (actually frame offsets of the CD) of the tracks, this means you must have all the tracks, they must be in correct order and your rips must be accurate to ensure track lengths they fall in the same second boundary as the original CD in order to be able to recreate the correct id.
3. The freedb algorithm is poor in that it allows two cds with same total length and same number of tracks but different individually different track lengths to have same id since the track length calculation is stored in a byte and can therefore only have 255 different values.
4. This makes freedb totally unsuitable for matching singles, e.p's.
5. Doing additional lookups to freedb will slow down matching.
Having said all that I am currently experimenting with a way of looking up discs from Albunack (MusicBrainz/Discogs) using an algorithm similar to freedb but without issue 3 and to a lesser extent issue 2.
This will find its way into next release of SongKong and then later on Jaikoz.
I expect I will take another look at freedb if it would help plug remaining gaps.