We noticed an issue with task creation that caused some requests to fail. While these failures were occurring there was a small increase in request latency. We found and addressed the immediate issue, and have a long-term fix being prepared for release soon.