I've been in the sport for a long time, and over the years have collected a substantial library of books about running and training.
Recently, I took at look at one called 1975 Marathon Handbook, which believe it or not includes a U.S. list of all sub-3 hour marathoners in 1974.
The 1975 Handbook includes an article that describes Ken Young's "Collapse Point Theory", which goes like this:
Training mileage over the previous 6-8 weeks sets the limit of how far one can hold a fast pace. That limit is about 3 times the daily average. After that point, the pace slows dramatically, and the runner may even have to stop.
In other words, a runner that averages 9 miles a day for 6-8 weeks has a collapse point of 27 miles.
Supporting the theory is a look at the old Oregon's Trail's End Marathon by researcher Paul Slovic. Generally, here is what he found:
1. Sub 3 hour runners averaged 9 miles a day
2. 3 to 3:30 hour runners averaged 6 miles a day
3. 3:31 to 4 hour runners averaged 5 miles a day
4. 4:01 and up runners averaged 4 miles a day
Further analysis showed that the average pace of Group One dropped 14% in the last 6 miles. Group Two slowed 22%, Group Three slowed 37%, and Group Four slowed down by 58% after the 20 mile mark.
Here is another way to look at it- if you average 40-50 miles a week, your collapse point will occur between 18-20 miles. You'll finish, but you are also likely to experience a significant slowdown the last 6 miles.
Clearly talent, pace, the number and length of long runs, environmental conditions, the course, and other factors play a role, but I wouldn't be surprised if this 70s Collapse Point Theory still applies to most marathon finishers today.