This is not only about merged questions from same OP, but also about normal duplicates from same OP (i.e. self-duplicate questions).
The merged ones are a particular case because they are locked posts and hence, not qualified for automatic deletion, besides not being eligible for voting, etc.
This will almost happen because OP asks a question, then nobody answers it (or answer in a way OP is not satisfied), then OP asks the same question again instead of editing/improving the original one. Those posts are not useful to have around because as written by the same OP, stub and master have nearly the same wording/problem approach, and are not useful for broadening the search scope (and consequently finding the canonical answer). Moreover, keeping them around sends a confuse/incorrect message it is ok to re-ask instead of editing.
With that in mind I wrote a query in Stack Exchange Data Explorer (SEDE*) to find unanswered self-duplicates which outputs the list of stub posts:
The unanswered aspect is because we can flag redundant posts for moderator attention and ask them to be purged without worrying about answers being deleted. We can also downvote such questions (the ones which are note merged) and push them towards automatic deletion.
Right now (jan/2019), the query is outputting a little bit more than 200 duplicate stubs.
I have finished reviewing the list of questions returned by the query, and with help of our moderators the list now has
18 14 questions (unanswered duplicate stubs). Some of them are already awaiting for the automatic deletion scripts to act (mostly the ones with negative score). Other (borderline) posts there could also go with one or two more votes, if anyone is interested in reviewing them.
Here are my personal findings:
- More than 90% of the unanswered self-duplicate stub list were just copies. OPs shouldn't have re-asked, but edited the original Q.
- I'd suggest moderators when merging self-duplicates, to evaluate the usefulness of the stub, and if not useful, delete it upon merging.
There is an additional query which is very similar to the previous one, but it finds self-duplicates in which the stub question has answer(s), and the target question does not. In such cases, the way to proceed is to reverse the linking relationship (stub becomes target, and target becomes stub) and evaluate if the new stub can be deleted (not useful). Here is the query, which is returning now (apr/19) 10 stub questions:
I wrote another similar query, but it finds answered self-duplicates (with answers both in stub and target questions), so they would need to be evaluated if a merge would fit answers from both questions, and then, deleting the stub. The query is returning 112 candidates for review (apr/2019). This is more tricky and difficult to review/accomplish, so I will leave it here for future reference (maybe discuss about it in another GIS Meta post).
*SEDE is updated in a weekly basis.