The places of greatest stress, and greatest likelihood of water infiltration, would be where the composite met the endcap (or a penetration, but you get extra longitudinal compression from the endcap). I'd expect microscopic water infiltration at an edge flaw to cause the outer layer to delaminate (maybe with a little crackling noise?). That makes the composite structure just a tiny bit thinner there, and exposes the next layer to the water. Maybe there's a flaw handy; maybe not. But I (and apparently quite a few others) would expect material fatigue after a while, and the next layer's boundary with the endcap to slowly fribble.
The picture on the right screams for context. Is that supposed to be a typical layer? Do they pull regular test samples? (If not, why not?)
FWIW, when the leak broke through, the water speed would have been about 270m/sec--end to end in 24msec. A blink is about 100msec. Leaks are bad things