The reason why T111189: Central Notice message groups are slow to index was created is that we sometimes want to re-populate data to restore the consistency, for example due to missed updates. The runtime of the script would be less relevant, if it worked differently by not deleting all data first (which causes user visible impact of translation memory/translation search not returning results), but if it would rather add current data first, and then delete what shouldn't be there.

This would require some kind of tracking what should or shouldn't be deleted, and it needs to be investigated if it can be done efficiently.