Avoidance of instability and congestion of task-oriented distributed sensor networks require a resource management scheme that is robust against time-varying network-induced delays caused by, for example, frequent and abrupt changes in the observed 'scene' in a highly dynamic environment. A convenient tool for its implementation is the so-called virtual per-flow framework which essentially decouples each incoming data link of a decision node. In this paper, the time-varying delays in each loop are accounted for by incorporating them as a time-varying F-norm bounded uncertainty. After converting the resulting uncertain discrete-time system into an equivalent scaled model without uncertainty, an LMI based static output feedback controller that establishes and maintains the allocated local resources uffer set-point and bandwidth at each processing/decision node - is presented. Simulation results demonstrate the effectiveness of the proposed strategy.