Been following this with my materials engineer hat on with some degree of dismay.
There's a reason really deep sea diving vehicles are spherical and largely made of steel. And not a cigar-shaped CFRP-titanium hybrid. The basic maths, engineering design, manufacturing and material properties are pretty self-explanatory.
My best bet is that it's one of two things.
a) propagation of failure at a stress raiser - either the viewport or where the ends bolt on to the sub.
b) catastrophic failure of the carbon fibre layup due to an air bubble / manufacturing defect.
The first is pretty well much what it says on the tin. Especially since we now know that the viewport wasn't rated to the depths the sub was operating at.
The second is less obvious to the layman, which is where a manufacturing defect in the layup - an air bubble - causes cracking in the internal structure. On a one-off dive it may well hold, but repeated i.e. cyclic loading will mean that the bubble keeps expanding and contracting as the pressure changes, increasing the size of the damaged area and weakening the structure as microscopic cracks and delaminations propagate through the matrix. And since you can't see it, you won't know it's there until it fails. CFRP doesn't fail in the same way as metal, and when it does, it just shatters, more often than not without prior warning.
Am a bit out of practice on the maths these days, but after seven years in academia spent doing various things with composites including smashing them up, I can still visualise the failure processes in my head.
I know it's a bit heartless in the way, but purely from an engineering perspective, the sub was an omnishambles that should never have been allowed off dry land.