The developers traced the issue to an auto-scaling issue in a part of our infrastructure, where a cluster wasn't properly adding more resources as quickly as needed in order to keep up performance under load. We've since tweaked the scaling configuration and it should not happen again. Our apologies for the slowdown!