Optimizing for one metric only, e.g., speed, leads to suboptimal outcomes, e.g., cost, scalability, etc.

I think while being fast, cerebra’s probably not very economical in fleets at scale.