I'm not sure that these intuitions are actually correct. For example, while it seems obvious that a four wheel solution should be faster and cheaper, that is only really true if you assume that software and controls algorithms are expensive. This robot needs fewer wheels, tires, frame structure, and drive motors than a conventional four wheel design, and the ability to lean into the direction of motion gives higher acceleration limits, increasing speed. Just like a bike is cheaper and higher performance than a recumbent tricycle, this might actually be cheaper and higher performance than something more conventional. If not today, Boston Dynamics seems to be betting that it will become so in the future.