This has been stable for over an hour now so we're marking this incident resolved. The issue was Apache2 hanging on a single container on a single host. As the API is anycasted by many containers on many hosts around the network, this only affected a small percentage of traffic from certain locations. Had Apache failed gracefully, others would have seamlessly taken the load and we'll continue investigating why it didn't offline.
Posted 11 months ago. Sep 21, 2017 - 12:44 UTC
A fix has been implemented and we are monitoring the results.
Posted 11 months ago. Sep 21, 2017 - 11:22 UTC
Our external monitoring is indicating intermittent API connectivity from certain locations. We're investigating.