I'd be curious about the math. It's really weighing the period of training and how often that training period is inflicted (turnover) against the productivity gains of overworking people.

E.g. if a business can force everyone to work through their half-hour lunch without pay, they're getting an extra 130 hours per employee per year for free. If it takes them a week (40 hours) to train a new employee, they can have 300% turnover for that role (replacing the worker in that role 3 times) and still net 10 free hours per year.

At 2 weeks of training, they net 50 free hours at 100% turnover and break even at 150% turnover.

At 3 week of training, they net 10 hours on 100% turnover, and would be better off letting people have lunch if turnover is higher than 100%.

Skipping lunch is a weird way to measure it, but it's hard to generalize "how much more productivity do companies get if they micromanage employees like Amazon does" or something like that.

It's only visible to upper management if people in the field aren't overworking themselves to meet deadlines impacted by someone leaving. Most problems are masked because people can't just "let it fail" and force mgmt to deal with it.

It's possible that the costs of high turnover might motivate a company to make work conditions better, but it's also possible that the company just finds ways to adapt to high turnover.