We investigate a model of random spatial permutations on two-dimensional tori, and establish that the joint distribution of large cycles is asymptotically given by the Poisson--Dirichlet distribution with parameter one. The asymmetry of the tori we consider leads to a spatial bias in the permutations, and this allows for a simple argument to deduce the existence of mesoscopic cycles. The main challenge is to leverage this mesoscopic structure to establish the existence and distribution of macroscopic cycles. We achieve this by a dynamical resampling argument in conjunction with a method developed by Schramm for the study of random transpositions on the complete graph. Our dynamical analysis implements generic heuristics for the occurrence of the Poisson--Dirichlet distribution in random spatial permutations, and hence may be of more general interest.