We study the Stochastic Traffic Engineering (STE) problem arising from the support of QoS-demanding live (e.g., real time) audio/video applications over unreliable IP-over-wireless access pipes. First, we recast the problem to be tackled in the form of a suitable nonlinear stochastic optimization problem, and then we develop a goodput analysis for the resulting IP-over-wireless pipe that points out the relative effects of fading-induced errors and congestion-induced packet’s losses. Second, we present an optimal resource-management policy that allows a joint scheduling of playin, transmit and playout rates. Salient features of the developed joint scheduling policy are that: i) it is self-adaptive; and, ii) it is able to implement reliable Constant Bit Rate (CBR) connections on the top of unreliable energy-limited wireless pipes.