This quote is a year old, forgive me for re-hashing it. I think it bears relevance to the entire discussion. I've put the important question in bold.The more spokes in a wheel the less reduction (%) the lowest spoke experiences. Will the magnitude of the cyclical range of tensile force in a spoke be related to fatique life? Probably (intuitively), but I don't know.
The spokes fatique and eventually, when @hobo gives that last push, the rear wheel spoke that happens to be weakest parts. On my ride last week each of my spokes (36 spoke front wheel) went through this cycle (1000N to >800N) over 100,000 times. Last year the spokes in my front wheel which I built it in Jan 18, completed this cycle over 5 million times. I used the original spokes from the fubarred wheel (rim) so I estimate these spokes have completed 20 million cycles with me aboard (plus no idea how many with the first owner on board). Pleased to say no front spokes broken yet.
The answer is yes, the cyclical range does have an effect on fatigue life (sometimes called Endurance Limit or even Fatigue Strength). Steel has a fatigue limit, which means that cyclical changes in stress have no effect on the fatigue life of the piece. In short, it means that if the changes are kept below a certain level, then the material effectively an infinite life. This limit ranges a bit and can be as high as 35% of tensile strength or as low as 65% of tensile strength.
We can relate back to spokes. A 1.8mm spoke has a tensile strength of about 2400N. Typically, we build to no more than 1500N and the cycles will be (big approximation here) about 200N. Even with a high limit of 35% you can see that a wheel can easily be built with infinite spoke life.
The important factors are: enough spokes; stiff enough rim and proper stress relieving.
Unfortunately, rims, made from aluminium, has a fatigue limit and it is pretty low too. Rims therefore crack long before spokes break, if the wheelbuilder knows his onions.
Edit: An error was pointed out to me and I've changed the magnitude of cyclical changes from 20N to 200N. The demonstration remains valid.