Two files that share four sequences are merged into one FASTA: the repeats are collapsed, with a count of how many went, and the records come out sorted by length. Duplicates are judged on the final sequence, after the transformations above.