Please please please change CommentClauseQueryFactory.generateIssueIdFromCommentQuery() to use a DocumentHitCollector rather than a for() loop that iterates over Hits.

We're still working on upgrading our Jira 3.13 to 4.1.2.
One thing I noticed was that comment queries were taking forever on my test server when compared to 3.13 running on the same server.
After a day of debugging and testing, I narrowed it down to looping through the Hits in CommentClauseQueryFactory.generateIssueIdFromCommentQuery().
The Hits class is deprecated and states,

 /** 
 * Iterating over all hits is generally not desirable and may be the source of
 * performance issues. If you need to iterate over many or all hits, consider
 * using the search method that takes a {@link HitCollector}.
 */

I modified our Jira to use a com.atlassian.jira.issue.statistics.util.DocumentHitCollector instead of a for() loop over Hits and our comment queries are now miraculously 6 to 8 times faster.

Here are the statistics for one of my tests before and after the change. For this test I ran comment ~ power which returns about 7000 issues. In Jira 3 I ran the equivalent by putting 'power' into the query box and checking only 'Comments'.

• Before the change
  • In Luke, the Lucene query took 3.5 seconds (and without the level restrictions it took 389 ms, which one should consider for another Improvement....)
  • In Jira, the Lucene query took the same amount time
  • Looping through the issue id's took 74 seconds
• After the change
  • Lucene query with hit collection takes 12s