Optimizing order of sets in MDX crossjoins

For scalar values a*b is the same as b*a. (although we saw that performance of these two could be different in MDX in “Performance of multiplication in MDX” article). But if a and b are sets, then obviously results are different – since the order of tuples in the resulting set will be different. But does it matter if we were to perform some other operation on top of result, which wouldn’t depend on order of the tuples ? Semantically – the answer is no, but there could be a big difference in performance. Let’s take the following example: