The specific set of circumstances that trigger this not being represented in their testing pool, including their public beta testing pool? I would expect that their public beta pool is pretty large, something hard to top with internal testing.
Maybe their testing procedures overlooked something that could have been included without increasing the testing pool significantly. But I rather think that fighting the underlying reason is an approach that is more likely to reduce such instances. That could start with having better programmers and better programming 'methods' that are able to predict and thus prevent circumstance that could result in such a failure. Maybe even the staggered release should be reconsidered, ie, if the 9.7" and 12.9" iPad Pro had been completely co-developed (and thus be as identical as possible internally) then it would be less likely that a bug affects only one of them (and if it affects both models, preventing and detecting it is easier).