This wasn’t what I was originally referring to but I still wouldn’t use it in training too frequently. One of the reasons is what Phoenix pointed out and the other is the very high neuromuscular (NM) demand. If you were just training for strength alone, training to failure (as we’re talking about it: 1-4 reps) has been proven by many of the best weightlifters (especially the bulgarians who max several times a day for several weeks at a time) as an excellent way to increase weight room numbers. If however, you’re also doing track / jump work, I think the NM demand would be too high to do frequently.
Correct me if I’m wrong here, but I believe Charlie has mentioned that if one of his athletes hit a new weight room max or PR on the track that he would drop their intensity down for a period of about 10 days. Note that in the case of weight room maxes hitting these new “maxes” doesn’t necessarily indicate what their true max was because they may have actually been able to go higher but if they exceeded previous bests they were stopped. This is a little more cautious than I’d be but I think Charlie has obviously got the right idea and I think this points to the effect that very high NM demand activities can have on the system and it’s future consequences on training and recovery.
Having said that, I did however use it occassionally (maybe two sets every 2-4 weeks) when I was at OU as a form of testing during intermediate points in the year when I didn’t want to do a full out max. I’d have them use 85+% of the old max and have them see how many reps they could do….basically your normal predicted max test but it would be in regular training day and not necessarily a test day.
All in all though, I think the research still points to multiple sub-maximal sets (even in low rep range) rather than a single set taken to failure. The greater volume of total reps that can be performed at a very high load (85+%) with multiple sub-maximal sets outweighs the fact that none of them are taken to failure.