Resolved
May 17 at 12:09am UTC
The source of the problem appears to have been that our API was unable to connect to one of its underlying data stores, most likely due to a networking interruption. This has recovered as of 00:02 UTC and traffic is being served normally once again. We will continue to monitor.