Apologies for a second post, but another great point of discussion around this topic: <a href="http://52weeksofux.com/post/646711369/cohort-analysis-measuring-engagement-over-time" rel="nofollow">http://52weeksofux.com/post/646711369/cohort-analysis-measur...</a>
I think the key is actually measuring the KPI's that matter. It's not total users its total active users. If you just report on total active users weekly, then you get to the same conclusion.
Quick note: your link to this HN discussion was invisible to me the first time I saw the page. It looks like a subheading that someone can very easily miss.