Hi @m.herrmann
Thank you very much for your answer.
In our implementation, we used SearchProvider, avoided sorting, and loaded the minimal number of system fields like id, key, status, type, summary from Lucene documents thus avoiding database queries at all. There is a perfect answer from Atlassian here - How to avoid Out Of Memory error using SearchProvider - #3 by kcichy
Where we stopped in our investigation is that pure Lucene search with loading minimal issue details takes 1-2 minutes for 10-15k issues in a Customer environment, while in our test environment it takes less than 1 second to find 40k issues.