Pressurisation is a big player.
As you pressurise a vessel - whether it's an aeroplane or a boiler, you get stresses forming in the skin which are a function of shape and pressure difference across the skin. They are called hoop stresses.
By introducing windows, which tend to be made of material that doesn't take tensile loads (hoops stresses are entirely tensile) the load which still has to go somewhere becomes concentrated around the windows. Bigger windows, or with sharp corners, tend to suffer bigger stress concentrations.
The Comet design teams didn't understand this well enough (it wouldn't be fair to say that they didn't understand it at-all). This isn't all that surprising since nobody until then had built large pressurised cabins to routinely fly to that sort of altitude.
So, cyclic hoop stresses, concentrating around the window corners, created large and unexpected fatigue damage areas. The fatigue cracks eventually joined up and the whole aeroplane unzipped with catastrophic consequences.
The DH design teams at the time had realised that their analysis was not as good as they would like, so had been happily testing a fuselage in a water tank to many times the number of cycles that the real aeroplanes (which broke up in the air) had been to. What they'd missed - which really was unknown at the time - was that by using the fuselage that they'd already used for limit strength tests, they were inadvertently using structure whose metallic/crystaline structure was modified by overload and had become much more fatigue tolerant than before. So the fatigue test fuselage was not representative of the flying aeroplanes.
A lot of lessons were learned from that, specifically...
- Use small windows which are as round as possible.
- Controlled overload of a structure tends to improve fatigue life.
- Non destructive testing of aircraft structure needed to be a lot better than it was. I've heard the Comet called the "Father of NDT", which isn't far wide of the mark.
- Be very very careful about the history of your test specimens, they need to be as close as possible to flight hardware.
G