Passive microfluidic sorting techniques based upon the interaction of particles with an optically defined potential energy landscape have possible advantages over active sorting techniques such as microfluorescence activated cell sorting (FACS), including ease of integration into lab-on-a-chip systems, reconfigurability, and scalability. Rather than analysing and deflecting a single-file stream of particles one by one, a passive approach intrinsically aimed at parallel processing may, ultimately, offer greater potential for high throughput. However attempts to sort many particles simultaneously in high density suspensions are inevitably limited by particle–particle interactions, which lead to a reduction in the efficiency of the sorting. In this paper we describe two different approaches aimed at reducing colloidal traffic flow problems. We find that continuous translation of the sorting lattice helps to reduce nearest neighbour particle spacing, providing promise for efficiency improvements in future high throughput applications, and that a flashing lattice yields a reduction in unwanted pile-up and spillover effects which otherwise limit the efficiency of sorting.