This issue occurred because there were thousands of records in the database that all had the same name.  Normal duplicate constituents have only a handful of potential matches, not hundreds or thousands.  These records do not have the properties of normal data as expected by the duplicate merge process, and as such, slow the process down significantly.

To successfully run the process, create a custom process to exclude constituents with atypical names or characteristics like this.