Nice! But that doesn’t answer the question. Do these optimizations don’t scale to multi-device workloads or not?