Sorry about this, I'm struggling a bit. Isn't the "load observed over the life of an aircraft" just a statistical distribution? Therefore a maximum doesn't make a lot of sense. It would make more sense to talk about standard deviations away from the mean unless the maximum was "maximum ever recorded in our dataset"
Suppose you have a bridge. You must design it so it can handle the maximum weight expected. You must not design it with a simple Gaussian distribution in mind.
For example, a small bridge in a seldom used road in the countryside might have a few cars per hour. That's a load of a couple of tons per car. But cars density isn't randomly distributed. If one is slow, then there might be a lineup of 4 cars in a row, and that's more likely then a simple calculation based on 1-(1-p)^4 where p is the single car density.
Then twice a day there's the school bus. And there's the occasional cattle truck, and semi, and anhydrous tanker. So you might have a school bus with a couple of heavy vehicles behind it, as a possible maximum design load.
The odds of this can't be described with a simple Gaussian, so standard deviations make no sense.
Of course, there's no need to go crazy overboard and design all bridges to handle a convoy of M1 Abrams tanks. That's why rural bridge might have a posted weight limit of, say, 10 tons, with only 1 truck allowed at a time. Even then, some people will push the limits, which is why there's a safety factor.
For example http://blogs.mcall.com/roadwarrior/2014/08/wehrs-mill-bridge... describes a wooden bridge which had a 10 ton limit, until some dumbass fuel tanker weighing 38.2 tons went through. Now the rated limit is 4 tons because of the structural damage.
So it's not "the maximum that was ever recorded in the dataset of all rural bridges" but "the maximum expected for this bridge", along with efforts to restrict higher loads.
First, I'm not doing structural design, so maybe someone will wander by with a better answer. Anyway, it's probably overly simplistic, but passing the load test would involve surviving the maximum expected load with zero damage. Implying that lesser loads would also never do damage.
In reality, cyclic fatigue is also considered.
This seems like a nice discussion that shows the complexity: