In reading one of the papers from the LTEE the concept of diminishing returns epistasis was mentioned, which also helps explain why it has to be physically true that the shape of the curve for the DFE of mutations must change over time as organisms become either more less well adapted.
Imagine an organism A that has some reproductive rate, and then a beneficial mutation occurs that increases it’s expected number of offspring by one. Now if this organism normally is expected to produce 100 offspring, and it has one more, it has improved it’s number of offspring by 1%.
Now imagine another organism B that has twice the reproductive rate that A had, and then this organism suffers the same mutation A did. It produces 200 offspring normally, and now can add one more. But now the mutation only has a 0.5% effect of improvement.
But that would imply the selection coefficient for this mutations has decreased. As the organism has gotten more fit, the effects of individual mutations have gotten smaller in proportion.
An analogy is that you are to push a heavy car that has run out of fuel to the nearest gas station. When you do it alone it’s very hard, but if you get another person to help, it’s half as hard, but then if you get one more, you’re doing a 3rd the work, and then with a fourth, a quarter the work. The gain from every additional person helping push the car becomes smaller and smaller, to the point of being neglible. When you’re 30 people pushing the car, you probably can’t even feel if any single person decides to take a break.