Personally I can’t suggest anything else here - I’ve not experienced these problems on any infrastructure including very large stacks. Perhaps @big-piano-35669 or @white-balloon-205 can help here?