The rubric was more of a "Y/N" for this question. But we'd refer to the author's solution as "great". Messier solutions would involve a lot of copy/pasta and occasionally inefficient or incorrect implementations that tried to repeatedly add.

I can't remember exactly, but I probably administered the question for Arthur when he interviewed (he may remember). I do (fondly) remember working with him.

A "solution" that repeatedly adds would seem to violate the atomic property, so I think I would call that invalid and not an actual solution?

The author of TFA mentions these solutions usually acknowledge the need for atomicity, and try to achieve it by reusing memcached's locking primitives. I understand the failure is that either they don't get them right or use up all their time without a complete solution.

To be honest: it's how I would have failed this question.