"The 3-week premarathon taper featured a 52% reduction in volume with no appreciable change in training frequency. This taper is congruent with the recommendations from a recent meta-analysis on the effects of tapering on performance, which found the ideal length of taper to be ~2–3 weeks, where training volume was decreased 41–60%, without any modification of training intensity or frequency (Bosquet, Montpetit, Arvisais, & Mujika, 2007)."
These guys went from running 142, 165, and 124 miles/wk to an average of 71 miles/wk in the last week including the race. Talk about a taper. If you're running 70/wk that would mean dropping down to ~35 miles (including race distance)... So, this data might not be particularly relevant to the sub-elite crowd, or those running significantly fewer miles - most of us here. And while these athletes have posted impressive times, this is only a case study from 2 training groups; coaches Dave Scott-Thomas and Richard Lee.
Regardless, I think another other interesting bit is that training impulse, a combination of physiological intensity, perceived exertion, and volume, peaked 7 weeks out from the race week. Unfortunately, we don't get the to see much into the specifics of their training.
The article also provides some interesting information on the manipulation of dietary carbohydrate and glycogen status in the athletes (though, muscle glycogen content was never directly assessed). This will be a focus for another day.