I found out that using 'Soundex' for the Title and 'Ignore' for the Author in an Author/Title duplicate search doesn't work well. With my library (~40K books) the plugin starts eating memory like mad, in the end crashing Calibre when it runs out, which happens in half a minute or less (this on a 2.4GHz Corei5 with 2GB RAM + the same VM)
It doesn't do this when using Soundex for both, or indeed any other combo I have tried (mostly Fuzzy/Fuzzy or Fuzzy/Ignore) or when using ISBN matching.
Those all work just fine, and have weeded out literally thousands of dups (probably close to 4000) from the mess that was my ebook collection