I'd say it's a mixture of chaos engineering and load testing. They know what performance they need for a safe flight, but they want to know how much margin of error they've got (and how close their simulations are), and what they can do to optimise.

It's a different apporach to say the Apollo program, where they did heavy up-front analysis, at the expense of cost-efficiency, speed, and innovation. They had one-shot for a flight, otherwise that's several $bn up in flames.

Even with the last few mishaps, it's an approach that seems to be working. If you look at Starship and Falcon's journey in comparison to SLS and Blue Origin, they have done so much in such a short timespan.