A tight bound on the speed-up through storage for quickest multi-commodity flows

Multi-commodity flows over time exhibit the non-intuitive property that letting flow wait can allow us to send flow faster overall. Fleischer and Skutella (IPCO~2002) show that the speed-up through storage is at most a factor of~$2$, and that there are instances where the speed-up is as large as a factor of~$4\mathrm{/}3$. We close this gap by presenting a family of instances for which the speed-up factor through storage converges to~$2$.