ML-DSA-44, unlike -65 and -87, can reject due to ||ct0||∞ ≥ γ2. It doesn't look like we have test vectors for that.
https://groups.google.com/a/list.nist.gov/g/pqc-forum/c/G8Zf0hC-uu0/m/Kb3qNJb0AwAJ has a very comprehensive set, unfortunately for Dilithium-2, but we should seek to reproduce them.
It would be good to generate boundary conditions for the other rejection cases, too. Especially for the ||r0||∞ ≥ γ2 − β check, since r0 is not used for anything else, so an off-by-one (or, in my case, a flipped sign of cs2) can go undetected.