Ageing - which is variably a function of humidity, UV, BVID / microcracking, and fickle fate, and to a lesser extent variability in manufacture, are surely the biggest headaches and the hardest to quantify. Hence the well known composite superfactor in AMC 619.
Every study I've ever seen showing tests to destruction of a range of components, particularly where the failure mode includes substructure buckling, shows massive variability in failure loads - even at start of life - sufficient to make the composite superfactor seem very non-conservative.
G