We've scaled up number of instances and we're no longer seeing an increase in errors.
Resolved
Identified
Due to a lack in GPUs available - we're seeing an increase in outages and a degradation in performance.