We present velocity and temperature field measurements for a 0.9 x 0.9 x 1.7 m glass tank in which two air jets mix and impinge upon the lid at ambient temperature and pressure. At jet Re approximate to 10,000, flow patterns below the lid were characterized for two inlet geometries: (1) "extended", in which inlet channels protrude above the tank base, and (2) "flush", a flat base without protrusions. This minor geometry variation produced distinct changes in the lid velocity field, appearing as three stagnant regions for the extended case and only one for flush. The dichotomy is attributed to system stability characteristics: jets are stable in the extended case and unstable for flush. In a separate set of nonisothermal tests, the impingement temperature field was measured for inlet temperature mismatches of 4 degrees C with jets again near Re = 10,000. A 50 m-long fiber optic distributed temperature sensor beneath the lid measured at 1350 locations. Like the velocity fields, the temperature fields differ for the two inlet geometries: good thermal mixing for the flush case and subdued mixing for the extended case. Simulations with the spectral element code Nek5000 replicated the observed stability dichotomy, duplicating the number of stagnant regions observed in the experiment and matching their locations within +/- 10 mm. Simulation data suggests that flush case instability is due to interactions between jets and wall flows at the bottom of the tank. The clear flow dichotomy exhibited by this two jet setup presents an unambiguous case to test the ability of CFD tools to predict subtle flow field changes driven by minor modifications in geometry in the context of thermal striping. (C) 2017 Elsevier Ltd. All rights reserved.