In supersonic flows, the separation in streamwise corners is a significant and widely encountered problem which can not be reliably predicted with the numerical methods commonly used in industry. The few previous studies on this topic have suggested conflicting corner flow topologies. Experiments of supersonic flow are typically conducted in wind tunnels with rectangular cross-sections, which use either a symmetric (full) or asymmetric (half-liner) nozzle configuration. However, the effect of the nozzle arrangement on the corner flow itself is not known. This paper examines the influence of nozzle geometry on the corner regions of a Mach 2.5 flow using a joint experimental-computational approach. The full setup and half-liner configuration are shown to produce different corner flow structures. The corner regions of the full setup and top corners of the half-liner exhibit thin sidewall boundary layers and a single primary vortex on the floor or ceiling. Meanwhile, the bottom corners of the half-liner configuration contain thick sidewall boundary layers and a counter-rotating vortex pair. Considerable vertical velocities are measured within the sidewall boundary layers. These are directed towards the tunnel centre-height for the full setup and downwards with the half-liner. The differences in sidewall cross flows between the two nozzle arrangements are likely due to distinct pressure distributions in the nozzle, where the secondary flows are set up. Measurements suggest that these nozzle-dependent transverse flows are responsible for the differences in corner flowfield between the two configurations. The proposed mechanism also explains observed differences in corner flow topology between previous studies in the literature; nozzle geometry therefore appears to be the dominant influence on corner flows in supersonic wind tunnels.