The hull imploded, not the window, but the shockwave probably pushed the window outward or shattered it, as it was only held strongly against moving inwards.

People make a big deal about the acrylic glass not being rated for the depth, but I think it's almost impossible that it was the window, and think it must have been the hull:

  * The properties acrylic are well understood, and the strength of an acrylic disc is trivial to calculate, even if not "rated" you can easily calculate if it's strong enough
  * The window should fail visually with a lot of warning before rupture
  * If the window failed with the titanium end cap, there would be no reason for implosion of the carbon fiber hull

The hull on the other hand is an irregular/non-homogenous part that is impossible to model accurately. There was no way to know how strong it was without extensive testing on a number of identical hulls- which they never did. The solidworks modeling method they used to give a ~2x margin of safety was nonsense as it ignores the realities of stress cycles, defects, etc.