Things like this generally make me wary. Not so much because of them causing side effects that are harder to disentangle, but their adoption indicating that the author is fairly stabilized, which is a giant side effect in itself.
It took me over a month to few "normal" again.
First week: headaches, back pain, vomiting, nausea, cramps, etc.
Second week: feeling like shit. Had no energy for anything.
Third week: a resemblance of being normal again.
Fourth week: feeling like I was at 60% of my previous energy levels and thinking "this sucks but if this is the new normal, I can get used to it".
Second month: energy levels are back, never feel sleepy again, focus is back completely, etc. I'm even thinking about gym, running, etc.
So this is my data point. From 3-4 expressos and of caffeinated sodas to zero.
I didn't stop taking sugar. I only dropped coffee and stopped taking caffeinated sodas... I didn't stop taking sodas.
https://en.wikipedia.org/wiki/Regression_toward_the_mean
It really is a very fascinating phenomenon. My friend who works as a psychologist have noticed that the most "successful" work is to always accept crisis patients immediately. If they have to queue for 2 months before getting clinical help, they have often improved by themselves, and it is therefore harder to reduce their problems further.
It's not that there's a regression-to-mean force that eliminates abnormal symptoms. It's more that the symptoms have some underlying cause that will fix itself over time.
If you start some intervention (such as stopping caffeine) at the same time you are naturally getting better, then you nay false attribute your getting better to the intervention rather than to its real cause: our bodies tend to bounch back from illness.