Running the ISBN "9780399159305" through the IdentifierLookupCoverageProvider creates a Work whose presentation edition with four contributors, all named "John Sandford".

"Presentation changed for Edition None (by John Sandford, John Sandford, John Sandford, John Sandford, pub=None, ident=ISBN/9780399159305, pwid=None, language=None, cover='https://covers.nypl.org/Content+Cafe/ISBN/9780399159305/Jacket.jpg')"

The problem is probably that http://classify.oclc.org/classify2/Classify?isbn=9780399159305&summary=false lists four different works, and each work has the same author, John Sandford.

The core problem is that we're building a number of Metadata objects but giving them all the same (ISBN) primary identifier. Each time you call apply() on one of these Metadata objects, it adds another John Sandford to the contributor list.

Rather than build a number of Metadata objects and applying them to the same Edition, we should build up a single metadata object and apply it to the ISBN Edition. This will let us make a decision to ignore further contributor information once we get some from OCLC.

We should probably not turn on the IdentifierLookupCoverageProvider until this problem is fixed – it has potential to do a lot of damage.