Even though JabRef has good duplicate detection, some manual de-duplication seems inevitable. Citations keys can help with this because a duplicated key represents a possible duplicate entry.
I can search for keys ending in a letter but this will reveal only the second duplicate and not the first, or else it will reveal all entries, depending on which option is selected in the preferences.
Here is a suggestion/idea for a fourth key-generator pattern:
Add letter A to duplicated key. Start on second duplicate key with letter B
Example library:
Carpenter2020
Carpenter2021a
Carpenter2021b
Search:
citationkey=\d[a-z]
Result:
Carpenter2021a
Carpenter2021b
I admit that this would not be a good approach for some use cases, because it would mean changing the value of existing keys over time when duplicates enter the picture. My use case, on the other hand, involves building new libraries for every project, which always requires a great deal of clean up, including correcting and regenerating citation keys.