Similar to proofs in computer science in general, I assume a great level of proof of robustness and correctness is required for scientific and mathematical visualization programs. How would one prove the correctness of a computer graphics algorithm or software, especially considering how our current understanding of any particular phenomenon is very likely to be incomplete?

Similar to proofs in computer science in general, I assume a great level of proof of robustness and correctness is required for scientific and mathematical visualization programs. How would one prove the correctness of a computer graphics algorithm or software, especially considering how our current understanding of any particular phenomenon is very likely to be incomplete?