Since you asked, I'll opine... The old school way was to design, build, and go drive it, and drive it, and drive it. That was found to be very expensive, and it required the test team to really understand how the customers used the vehicle and replicate that to the point that enough miles could be put on the machine to represent a lifetime, but within a very short (2-3 month) timeframe for that model year. The new school way is to design, build, and test it in a laboratory. It can be less expensive, but you now have to figure out how to apply loads in a laboratory that match those in the real world. You still have to understand how the customers use the vehicle, but now you have to figure out how to replicate that in a laboratory.
My career started in the old school way and transitioned to the new school way. Both ways can work, but both require that you develop a test that represents "real-world" use of the product. That's automatically easier to do in old school way, and with both ways, it's the hardest part of the entire process. In the new school way, that's called "developing a load case," which can be done from the vehicle level (using "multi-post rig"), the assembly level, and all the way down to the component level. The trouble is, the loads on a vehicle vary dramatically depending on who's driving, and in what conditions. As an engineer, if I know exactly what you're going to do with the vehicle, I can design it to survive for the expected lifetime of the vehicle, if survival of the vehicle is the only design constraint (it never is), and the expected lifetime of the vehicle is never exceeded (nearly impossible). Of course, as soon as that vehicle gets to thousands of customers, new ways to destroy it pop up, or the twelfth owner exceeds the expected lifetime, and suddenly the engineers are idiots.
As you said, "there is always a more worse, worst case." Designing an off-road vehicle to be used by the general public is very challenging. Here's one tiny example that will forever be in my memory bank:
We had a small aluminum part in the rear suspension of a snow machine that never, ever failed during testing. After we went to production, the first machines went to some cross-country race teams in Alaska because their season started the earliest. Within weeks of them receiving the machines and putting miles on them, every single part they had failed and very quickly. We scratched our heads and wondered what the heck happened? Were the parts defective? No, it ended up being a case of "bad load case development." When those Alaskan racers start their season, they ride wherever they can just to put miles on. In this case, that was on frozen rivers at high speed because there's not enough snow on land to cover rocks, logs, etc. that will destroy a snow machine. The frozen rivers become like smooth, paved race tracks to them.
That type of usage caused the part to be subjected to a high frequency oscillating load, and since all aluminum alloys are subject to a fatigue cycle limit (unlike steel, which, if the stress is low enough, won't break), the part saw more fatigue cycles in a few weeks of their testing than in a lifetime of service to other customers. We sent a couple guys up there to instrument the part with strain gauges and collect data on the loads it was subject to, then we built some beefier versions and quickly tested them in a lab under those conditions, picked the optimum version, and sent them to Alaska for testing - all before their season started. Those beefier parts never made it to any production machines because no one else road like that. We never had any other known field failures because no one else duplicated their load case. We just made them available to those teams so that they could run them on their race machines.