Could I suggest that when searching for exact matches by content, that the CRCs are computed lazily. i.e. it is only when a pair or more of files are found with exactly the same length that the CRCs of all files in the set need to be computed.
For any file that has a unique length, it should not be necessary to compute its CRC.
This would give an estimated performance improvement of ~ 100x when scanning my music library.
thanks for a great app!