Honest question: Have you worked with fluid simulations before? We cannot fully predict turbulent flow from first principles. I mean, we can, but when constrai

Honest question: Have you worked with fluid simulations before? We cannot fully predict turbulent flow from first principles. I mean, we can, but when constrained by real-world computing hardware, we can only do it in very tiny domains. Turbulence requires way too many orders of magnitude in spatial resolution to fully resolve it in "real world" problems. Therefore, any practical simulation for engineering purposes involves some way of modelling (read: approximating) turbulent behavior at sub-grid scales. There are tons of these models suitable for all kinds of different applications, domains, flows, etc. - but they are all just models. Part of the engineering effort is finding the right model for the task on hand, and for that you need experimental data. This of course doesn't apply only to fluid dynamics (but also e.g. the overall dynamic behavior of the entire assembled rocket and rocket engine, which could interact with the fluid flows in a nontrivial manner), but the point is that simulations aren't automatically accurate because you do more of them or throw more money or compute at them (that is also a good way to burn cash) - they are still approximations of unknown accuracy until you actually compare them with data from reality to confirm whatever modeling assumptions you made. The "least cash expenditure" approach will thus involve both simulations and experiments, and you will eventually reach the point where putting money into making rockets (potentially) crash will give you better lessons than running yet another simulation with (unbeknownst to you) subtly flawed assumptions.