That's not how it works. The demand for engineering hours is an order of magnitude higher than the supply for any given game, you have to pick and choose your battles because there's always much, much more to do. It's not bizarre that nobody verified texture storage was being done in an optimal way at launch, without sacrificing load times at the altar or visual fidelity, particularly given the state the rest of the game was in. Who the hell has time to do that when there are crashes abound and the network stack has to be rewritten at a moments notice?

Gamedev is very different from other domains, being in the 90th percentile for complexity and codebase size, and the 99th percentile for structural instability. It's a foregone conclusion that you will rewrite huge chunks of your massive codebase many, many times within a single year to accomidate changing design choices, or if you're lucky, to improve an abstraction. Not every team gets so lucky on every project. Launch deadlines are hit when there's a huge backlog of additional stuff to do, sitting atop a mountain of cut features.

> It's not bizarre that nobody verified texture storage was being done in an optimal way at launch

The inverse, however, is bizarre. That they spent potentially quite a bit of engineering effort implementing the (extremely non-optimal) system that duplicates all the assets half a dozen time to potentially save precious seconds on spinning rust - all without validating it was worth implementing in the first place.