I can't find a primary source, but as per Layne Norton regarding his PHAT routine:
Why aren’t all sets to failure?
Failure is a tool and has to be used correctly. During the first 3-6 weeks of the routine I would NOT recommend taking sets to failure as doing so will burn you out physically and mentally very fast. I’d recommend stopping 1-2 reps shy of failure. Once you get adjusted to the volume and frequency then you can start adding in failure for your power movements and some of your accessory/auxiliary work as your body gets more adjusted. You should never train to failure consistently more than 6 weeks in a row without at least a partial break from it. The reason I recommend this is because if you constantly train to failure it will decrease your performance, strength, reduce the volume you are able to tolerate, and ultimately reduce your hypertrophic capacity. There is this notion out there that any set not taken to failure is a wasted set, but that’s complete NONSENSE that has been perpetuated over the years by people who have not read the research.
Indeed, research has shown that sets taken near, but not to failure are almost, if not as effective as sets taken to failure on inducing growth and there is the added benefit that they do not overtax your nervous system to the point where it reduces your strength, power, and volume output.
For example, if you train to all out maximum concentric failure on bench press and hit failure on your 6th rep. The next set you probably will only get 3-4 reps, and the following set you might get 1-3 reps. As you can see your output decreases very quickly. But if you only took the first set to the 5th rep, it is quite likely you could take each subsequent set to 4-5 reps and so overall you have actually had a more effective workout by staying away from failure because you were able to maintain your power/strength output over several sets.
People don’t realize that overload is cumulative over the workout and adding more volume is actually a way to induce more overload, not just by adding more weight/reps. Again, there is nothing wrong with taking some sets to failure once you are adapted to the routine, but it has to be properly periodized to avoid performance decrements and CNS overtaxing.
Couple this with no quantification on how much the spotter is really
lifting and training beyond failure is just not sustainable. If you're going that way, you should be better off doing drop sets or working some parts in isolation or utilizing chains or something.
But, hey, the proof of pudding is in the eating. If it works despite common knowledge so be it.