One thing I didn't mention in the article is that we did allow some people to run their code. When we did do that, 90% of the people freaked out when they saw that their output didn't match and just froze. I think it was the interview nervousness/pressure.
They couldn't solve it even after we told them to take a sip of water, relax, take a breath and take their time.
So, I think questions like this might just not be a good fit to ask in an interview whether we allow people to run their code or not. I guess it's because these questions are so new to do out of the blue if you don't deal with them everyday.