[testplzignore]

Strange thing about the duration of this outage... From logs I have, it seems to have lasted exactly one hour, from 15:38 to 16:38. Their Twitter account also said "disruption lasted up to an hour", though they incorrectly said it started at 15:46 (did it take 8 minutes for their monitoring to notice?).

That makes me think that whatever the fix was, it had to wait for some one-hour cache to expire before it took effect. I'm very interested to find out what the cache issue was, more so than what the original bug was.