The issue is a lot of teetotalers don’t drink anything because of their existing health conditions, really bad obesity, hypertension, liver problems, etc. So those that don’t drink at all are actually less healthy than the average population, and those that drink in moderation are obviously healthier than those who drink a lot. So the results look like moderate drinking is the most healthy but there’s an (or a lot of) omitted variable bias.
Shhhh you just described iterative development. Careful not to be pro agile, or the developers with no social skills will start attacking you for being a scrum master in disguise!