> Why are people assuming this? How about a leak that fills the sub with water? Is a leak impossible to keep stable at those depths?

Part of this is the use of carbon fibre, which is a very brittle material. In bike frame failures, for example, carbon fibre doesn't tend to rip a little or bend, it often fractures greatly or even shatters.

Add the enormous pressure difference, and this is why most people assume instant catastrophic failure.