The issue with the performance of model servers has been identified and a workaround has been implemented. We are still investigating the root cause.
