You are conflating hours with productivity.

Let's assume that no engineer is getting less than 2 weeks of vacation off a year on paper. Most of these articles are written by people giving 4 weeks of vacation. So we're talking about the difference between a 'stingy' vacation policy and a generous vacation policy being 2/52 weeks in the year.

If people gain more than 4% productivity by avoiding burnout, you net win by giving people more vacation. I would argue that net productivity gain from letting people de-stress is way more than 4%, though I have no numbers to back that up.