In this paper, we analyze the network properties of the Italian e-MID data based on overnight loans during the period 1999-2010. We show that the networks appear to be random at the daily level, but contain significant non-random structure for longer aggregation periods. In this sense, the daily networks cannot be considered as being representative for the underlying `latent' network. Rather, the development of various network statistics under time aggregation points toward strong non-random determinants of link formation. We also identify the global financial crisis as a significant structural break for many network measures.