The longer we wait to physically test a structure the more likely we have field failures from actual use that drive changes, so the physical test of the original design becomes less meaningful.
In the balance between using standardized tests and random field time, there is a trade off. A test gives you information much sooner and in greater detail about the weaknesses. Yet it is the same test regardless of the corner conditions where the machine may actually be operating. Another benefit of standardized testing is that we know what operation was most destructive. However, with actual use in the field, it is much harder to know what loads and boundary conditions caused the failure. It's like Apollo 1, in some ways it saved other astronauts lives since it failed during testing instead of a mishap like Apollo 13, where it would have been nearly impossible to understand the root cause, and might have happened multiple times.
The point is, a test can give you targeted information about one component's response to the loads and boundary conditions applied. Actual use on the other hand gives you information about the entire variety of loads and boundary conditions that will be applied over the entire machine. While many failures are of simple components, others are due to systems of components. It is a balancing act between standardized testing and field testing.
While this is strictly engineering I am talking about, testing a mountain bicycle on the same one mile loop you always test on, versus any variety of trails and other surfaces with serious riders, it could apply to many things.