A spike in traffic seems to have resulted in a spike in response time. I also see H27s.
https://mastodon.zunda.ninja/media/1l5mIgI92JUvSnCqq7E
The spike in traffic was mostly HEAD request to /users/zundan/updates/533 as well as GET to /users/zundan/updates/533.atom from different IP addresses. I suppose the conversation was picked up somewhere and, well, maybe clients rendered the conversation on the fly on their browser?