When does a decohering quantum system become classical? Generalised noncontextuality gives one criterion, for scenarios rich enough in preparations: a system is classical once its statistics admit an underlying model in which operationally indistinguishable procedures are physically identical. Where it crosses depends not on the state and channel alone but on which preparations, transformations and measurements the experiment offers, and if every measurement is allowed there is no crossing. For any pairwise unbiased bases in any dimension, given preparations holding their eigenstates and surrounding the maximally mixed state, the question reduces to a finite linear program; for a qudit of odd prime dimension $d$ restricted to the stabiliser subtheory, with one depolarised magic state, we solve it in closed form. With preparations and measurements alone the qutrit strange state is classical at zero noise; with the Clifford gates added, each channel represented once, it crosses at 3/5; and with a random gate represented, as Spekkens’ treatment asks, by the matching mixture of stochastic maps, at 3/4. A Clifford orbit of assignments supports such a representation only if it has $d^2$ members, leaving, in every odd prime dimension, only Gross’s discrete Wigner function: the crossing sits where it stops being negative.