Numerical climate model simulations suggest that global warming is enhanced or hampered by the spatial pattern of the warming itself. This phenomenon is known as the ``pattern effect'' and has in recent years become the most promising explanation for the change over time of climate sensitivity in climate models. Under historical global warming, different patterns of surface-temperature change have emerged, notably a yet unexplained cooling in the Southern Ocean and the East Pacific. Historical climate model simulations notoriously fail to reproduce this cooling, which may contribute to the deviation of the simulated global-mean warming from the observed record.Here we qualitatively investigate the potential impact of historical and other surface-temperature pattern changes by changing the ocean heat transport convergence (Q-flux) in a slab-ocean model. The Q-flux changes are always implemented such that in the global mean they impose no net forcing. Consistent with earlier studies we find that the impact of a negative Q-flux change in the Southern Ocean has a stronger effect than in other regions because of a feedback loop between sea-surface temperatures (SSTs) and clouds in the Southern Ocean and the stably stratified regions in the tropics. The SST-cloud feedback loop facilitates the expansion of the Antarctic sea ice, indeed taking the model into a Snowball-Earth state. The intensity of this effect is found to be model dependent, especially due to differences in the cloud parametrisation. In experiments with deactivated sea ice the impact of the negative Q-flux change is much weaker.