Durability needs a better discussion on what happens during failure

This fails to mention that when a failure occurs while using durability requirements, the operation may or may not have been successful and that it's up to the application to determine what to do next in whatever is application specific.

I'd also argue it's a bit verbose in that it spends a lot of time talking about system implementation rather than the interface and semantics. Durability requirements have non-obvious semantics because it's a function that appears atomic, but is built out of side effects that can fail in a number of ways. These need to be a lot clearer.