Service availability in wireless networks is highly dependent on efficient resource allocation and guaranteed Quality of Service (QoS) amid overloads and failures. This paper addresses optimal bandwidth allocation in a hybrid network (cellular and ad hoc), where added reach through an ad hoc overlay is combined with the stability and essential services of a cellular network. The paper builds on a near optimal approach in which Resource-Utility functions are used as a means of adaptive delivery of QoS, user differentiation, and maximisation of system level utility. It distinguishes between non-adaptive, semi-adaptive, and fully adaptive applications. First, the global cellular bandwidth allocation (in the presence of multiple routes through ad hoc relays) is cast in terms of a Linear Programming problem. Second, a heuristic algorithm that has far lower computational overhead and accrues at worse 12% less than the utility of the optimal solution is presented. Both algorithms are implemented within a model of a hybrid network on top of the J-Sim simulation environment. Comparative studies are made to show effective load balancing and crash tolerance in the presence of a high traffic overload.