There is no real evidence that training to failure provides optimal results as opposed to training to near-failure (as SF does).
^This is a pretty good summary.
From the article:
"The hallmarks of successful training are long-term consistency and progression. But progression must be gradual- very gradual- if it is to be consistent. Many athletes insist on always taking a set to utter failure, even if it's not necessary to achieve a new personal record. But these same athletes neglect to project these gains into the future, which reveals the impossibility of continuing these gains. As an example, if you manage to put 5 pounds a week on your squat, this equates to 20 pounds a month, and 240 pounds a year. If this could be continued for even three years, you would be a national level powerlifter, with size to go along with it! A better approach is to achieve very small increases in load on a regular basis, even though you won't reach failure. These smaller increases are easier for the body to adapt to, and recuperate from. Taking each and every set to complete failure is like trying to run a marathon at sprint speed- after a very short period of sprinting, you'll have to slow down considerably, if you expect to finish the race."
A few more articles.
It seems training to failure constantly greatly increases the chance of over-training. The theory I ascribe to is that it is better to increase your workload by doing a great number of reps in small sets and building from there, rather than blowing out your system with as many as you can do until failure.
For me, fitness is a life-long goal, and training to failure makes me hate working out. I have to make smaller steps than a lot of the folks here to keep me consistantly working out, but that is just what works for me.