I will talk about my result that a finitely generated nilpotent group is Frobenius stable if and only if it is virtually cyclic. I will explain how the non-stability comes from a 2-cohomology class of a particular form that allows one to build an "efficient" sequence of projective representations. From there the "winding number argument" due to Kazhdan, Exel, and Loring is used to show that these projective representations are far from genuine representations.
