We’re experiencing massive internal network problems at the primary Flickr data center. You may notice this with occasional pageviews (or, at times, most pageviews) taking *much* longer to come across the wire. Oddly, they either take the usual 0.2-0.5 seconds per page or exactly 3, 5, 9 or 21 seconds. We have engineers and network ops people running around like howler monkeys trying to diagnose and fix this, but in the meantime, we all suffer.

UPDATE: This problem was resolved yesterday afternoon shortly after this posting (turned out to be that we had the do_not_work flag set on some of the database clusters). Sorry for forgetting to update ;)