Auto-fix duplicates of questions deleted by the Roomba

Question

While normal users can no longer vote to delete questions that have other questions closed as duplicates of them, the automatic deletion script (a.k.a. the Roomba) still can, and does, delete such questions. This leaves the closed dupes permanently pointing to a 404 error page, which is surely not a good user experience for anyone who happens to stumble across them.

(In case you'd like to see some examples, here's an SEDE query to find such orphaned dupes.)

I think something should be done to fix this. I can think of a number of possible solutions (some of which I've already suggested earlier), such as:

1. Send the duplicates to the reopen review queue, but with "Delete" / "Recommend Deletion" shown in the options instead of "Leave Closed" (since it makes no sense to actually leave such questions closed but not deleted).

2. Send the dupes to the normal reopen review queue, and raise an automatic ♦ mod flag if the outcome is "Leave Closed".

3. Same as above, except that, if the dupes would be eligible for automatic deletion if they were normal closed questions, just delete them immediately instead of sending them to review.

4. Same as #2 or #3 above, except that any dupes with a negative score (that are not eligible for immediate auto-deletion, if implemented) would be sent to the LQP queue instead (with a mod flag or an automatic reopen flag if the outcome is "Looks OK").

5. Automatically re-close the dupes using the same reason as the deleted question was closed with. This gets rid of the 404 dupe link, and in some cases may also make those questions, in turn, eligible for automatic deletion. (If the deleted question was not closed, either automatically reopen the dupes or send them to review as above. IME, this should not happen very often.)

6. Always raise a ♦ mod flag any time a question with dupes is deleted for any reason, and just let the mods deal with it. (As long as it doesn't happen too often, this might be the simplest option.)

In any case, IMO, it doesn't really matter how these orphaned dupes are handled — the important thing is that they should be handled somehow, so that they won't stay closed as duplicates of a question that doesn't exist.

---

Related feature requests:

• Change to operation of Roomba when deleting duplicates: Proposes that "the Roomba [sh]ould still go ahead and delete when other criteria are met and the question is a duplicate of another by the same user." This would fix some orphaned dupes, but not all. My suggestions #3 and #5 above could be seen as extensions of this proposal.

• Don't let an answered question be closed as a duplicate of an unanswered one, even if they're by the same user: This feature-req, which I just posted, attacks the issue from the other direction, and should make orphaned dupes less common (but not impossible) if implemented.

Answer 1

There are three kinds of questions eligible to be roomba'd: downvoted unanswered questions, old low-activity unanswered questions, and non-upvoted closed questions without upvoted answers.

The ultimate goal, of course, is for people to find answers to their question, whether that be by finding a duplicate with the answer or by waiting for a reopened question to be answered. As seen in the roomba'ing criteria, it's highly unlikely that the now-deleted question had a good answer. When a dupe target is deleted, we want to maximize the chance of someone being pointed at the right answer.

So, I propose the following:

• When a dupe target is deleted, the questions closed to it are all automatically reopened. This allows new answers to appear while things are worked out.
• The system selects the new best potential dupe target from among the newly-reopened questions:
  • The one containing the answer with the highest score (tiebroken by question score and then by oldest question).
  • Or, if there are no viable answers (scored at zero or more), the question with the highest score (tiebroken again by age). Good answers generally come in response to good questions.
• The non-selected questions are all dumped into the Close Votes queue as possible duplicates of the best potential target. Once the reviews are complete, deduplication will have been again achieved.

The most apparent downside to this scheme is that the question with the most upvoted answer might not necessarily be the most general case. An automatic comment from Community on the best potential target could include links to the other candidates.

Also note that a question can be in the Close Votes queue as a possible duplicate of multiple questions, so the top two (or more) candidates for best dupe target could be shown to reviewers, should that be judged helpful.

If a really epic answer appears on a question that isn't the best possible dupe target, reviewers can vote to close in the opposite direction (with a tiny bit more effort) after seeing it in the appropriate tab of the review interface.

It might be wise to have some upper limit on the number of questions that can be automatically reopened. Though it probably wouldn't happen very often thanks to the deletion criteria, a flood of two dozen very similar questions risks spreading any new answers very thin. In such cases, an automatic mod flag might be more appropriate.