Between Apr 21–22, we had some capacity issue in our workflow system that powers our orchestration layer. In between incident identification to recovery by adding more resources, we dropped some API calls intermittently.