It's annoying that there isn't enough research on calorie restricted diets to conclusively show the benefits and drawbacks.
All this said, fasting is an extra stressor on the body, I think... something to be aware of.
Autophagy can also be achieved by limiting protein over a certain time window, and some have advocated this as a likely preventative for diseases like Parkinson's and Alzehiemer's, involving "plaques" that are normally cleaned up during authophagy.
To get good data you really need to monitor people 24 hours a day.
Unfortunately people love their privacy too much to make that happen.
I believe the problem isn't with privacy but just that it is difficult to keep notes of what you eat daily. Make a product that makes monitoring what you eat very easy to do and you can get more data that's solid. What Apple did with HealthKit did in fact help, we just need more of that. I know a buddy who lost weight easily because of a smartwatch (Apple Watch) that he uses to monitor his input because it is so easy and quick.
Ideally, you would only eat what you (or a coach, or an app) planned to eat. Most people should have a nutrition routine where they eat the same thing every day or week. Eating for pleasure should be the exception, not the norm. Tracking only the 5% of food that's not part of your routine would be a lot easier (especially if packaged and provides a bar code).
All the prisoners would need to be separated and lose access to the commissary. They would probably be willing to trade their commissary for access to an XBox in their cell or something similar.
Then we can control exactly what they eat, and measure the effects of a calorie surplus or deficit over time. We could even give prisoners the same amount of calories but different macros, to finally prove the effects of a long term ketogenic diet are different than just calorie restriction.
The food would have to be more palatable than what they receive now, and they would have to agree to consume 100% of it and not try and trade with other prisoners. This probably involves a guard who watches them eat in their cells. It would be costly, but not nearly as expensive as the controlled environment studies in a hospital (metabolic wards) that Gary Taubes has done before.
it's like trying to train a nn by providing 95% random noise with input and not bothering to understand 95% of output.
seems like taking everything into account is just too expensive.
That is basically what most people would love to see.
Unfortunately, studies like this are notoriously difficult to run due to terrible self reporting. To complete an accurate study like this would likely take constant monitoring and an administered diet for the participants.