Yesterday, a small portion of our API requests began failing with 500-level responses and/or connection timeouts between 18:42 – 19:37 UTC and once more between 21:12 – 21:16 UTC. We determined the root cause for this failure was a performance degradation that occurred within our hosting provider's block storage service for one of our availability zones.
Our engineering team is reviewing the details of this incident and will be making monitoring, configuration and system changes that will ensure that failures of this nature are less impactful in the future.