I just worked with Jaikoz and they store some time consuming information like accoustic fingerprints and Musicbrainz ID into the file.
The next time it is not required to calculate that stuff again.
Also maybe Similarity coud utilize already stored IDs OR if they are sufficient in quality use them for precise comparison instead.
Also a cooperation with e.g. Jaikoz may be interesting as YOU make the best duplettes checking :-) !
