A common scenario: in the development environment, a table’s ID column is missing its unique constraint, and developers have been creating their own data for testing. Suddenly someone realizes that the ID column has duplicates. The DBA is summoned. "Enable the unique key", the DBA is told, "but do not delete our data."
Can the DBA escape this catch-22? For sure, says this post — with an approach of replacing *only* the duplicate IDs with unique values, and letting everything else stay untouched.