I was havin a peak usage of scanning 900 objects every minute. So searching for this was fairly easy. Just look into the access log what in low traffic times gets to the server, and it was the rss feeds.
They queried the most current objects, but only 30 of them. But: The index was only over date created, not date updated which I used for sorting downwards.
Adding the index in the source automatically created it in the database after deployment. That CI stack really is a help.
The mini diff: