I couldn't figure out why a caching bug would likely be restricted to just the period where the key was published with a revoke bit set. So I decided to re-try the experiment I did in the airport, but this time with data captures and "real science". The results contain graphs, so rather than write it up here, I'll direct you over to my write up on this web site: https://www.isi.edu/~hardaker/news/20190404-bind-bug.html TL;DR: some versions of bind (at least), in certain configurations, still sometimes (15% of the time) enter a state of sending a large number of DNSKEY queries on a regular basis. Today. -- Wes Hardaker USC/ISI