We investigate some model and proof theoretic aspects of sabotage modal logic. The first contribution is to prove a characterization theorem for sabotage modal logic as the fragment of first-order logic which is invariant with respect to a suitably defined notion of bisimulation (called sabotage bisimulation). The second contribution is to provide a sound and complete tableau method for sabotage modal logic. We also chart a number of open research questions concerning sabotage modal logic, aiming at integrating it within the current landscape of logics of model update.