Earlier studies have conjectured that redshift drift is described by spatially averaged quantities and thus becomes positive if the average expansion of the Universe accelerates. This conclusion is reevaluated here by considering exact light propagation in a simple toy-model with average accelerated expansion. The toy-model and light propagation setup is explicitly designed for concordance between spatial averages and averages along light rays. While it is verified that redshift-distance relations are well described by average quantities in this setup, it is found that the redshift drift is not. Specifically, the redshift drift is negative despite the on-average late-time accelerated expansion of the model. This result implies that measuring redshift drift signals at low redshifts gives the potential for directly falsifying the backreaction conjecture. However, the results are based on a toy-model so it is in principle possible that the result is an artifact and that redshift drift is in reality well described by spatially averaged quantities. The result therefore highlights the importance of developing exact solutions to the Einstein equations which exhibit average accelerated expansion without local expansion so that the relation between spatial averages and observations can be firmly established.