They do run stress tests, alright. But those are not representative of final performance, because a lot of optimisation and reengineering happens between first dipping your toes into a machine and releasing your first game on it.
Last of Us remaster, for example, was still running at 30fps until a couple months before it actually went gold.
No actual developer would say "we ported our code and it runs at 4k60 with dips" because that's not how a dev thinks. Thats how a consumer sees it from digital foundry. A dev does their test to find new bottlenecks and potential strenghths to lean on.
He said he made a similar test to spiderman's. Ok, was it with an actual game's data? Was it a naive port of the game, or did they refactor how the data was packed? Did they remove duplicates, changed layout, block sizes, experiment with different compression methods? Because that's the kind of thing an actual dev would be doing and those are the info he would be interested about and exited to convey in a hypothetical leak. Yet, no information about that on this leak, but a handwavy "30% performance increase until highest res restures are enabled" which betrays a completely consumer-like perspective on that stuff.