This article presents a comparative stability and performance analysis of control architectures for bimanual kinesthetic telepresence systems. Asymptotic stability of the kinesthetic closed loop is investigated considering analytical models for the dynamics of display, manipulator, operator, and environment. Stability bounds are computed for typical environmental properties by use of the parameter space approach. In addition, the quality of kinesthetic feedback is evaluated by introducing a novel quantitative performance measure. The application of both methodological approaches to typical two-channel control architectures demonstrates, that none of these can provide high kinesthetic feedback quality and robust stability for arbitrary environmental conditions. As a consequence, a structure adapting control architecture is proposed, which adapts the algorithms for local display and manipulator control, depending on identified environmental conditions. An extended stability analysis is presented for this architecture, and its appropriateness for bimanual kinesthetic interaction in remote environment is discussed. The usability and efficiency of the proposed approach are tested by experiments with a developed telepresence system applied to teleoperated demining.