The inherent complexity characterizing production and/or service networks strongly favors decentralized and self-organizing mechanisms to regulate the flows of matter and information in circulation. This basic observation motivates us to study the flow dynamics in queueing networks roamed by autonomous agents which, at a given time and at a given vertex location, select their routing according to (individual) historical data (such as waiting times) collected during their past progression in the network. For several simple network configurations and despite the intrinsically non-Markovian character of the dynamics, we are able to discuss analytically the emerging collective dynamics that such a circulation of autonomous agents generates. Feedback loops in the network topology coupled with the presence of delays in the routing selection mechanisms produce a wealth of dynamical phenomena like self-sustained generically stable oscillations, spatio-temporal patterns, stabilization by noise phenomena and oscillator synchronization that are explicitly discussed in this paper.