We use a unique dataset of transactions from the real-time gross settlement system TARGET2 to analyze the behavior of banks with respect to the settlement of interbank claims. We focus on the time that passes between a payment’s introduction to the system and its settlement, the so-called payment delay. Delays represent the means by which some participants could free ride on the liquidity of others. These delays are important in that they can propagate other delays, thus prompting concerns that they could cause system gridlock. This paper characterizes the delays in the TARGET2 and analyzes whether delays in incoming transactions could cause delays in outgoing transactions. We distinguish between the potentially mechanical pass-through of delays and the reaction of one bank to its delaying counterparty, and we propose a set of instruments to tackle endogeneity issues. We find evidence that delays do propagate downstream; however, in most cases the effect is rather limited. As for delaying strategies on a payment-by-payment basis, contrary to the theoretical literature, the data show only very weak evidence. This conclusion opens a venue for research how banks may rather follow persistent liquidity management routines.