A key component of successful stream restoration is to re-build resilience to the prevailing disturbance regime. In undamaged streams biotic resilience comes substantially from the use of refugia during both floods and droughts. For streams with channels heavily damaged by human impacts, such as excessive sedimentation or wood removal, biotic resilience to drying may be restored by the provision of refugial pools. With severe drought, for example the current one in Australia (1998--?), stream restoration becomes difficult. Drought in reducing natural refugia can lead to local extinctions together with the loss of putative reference sites against which a successful restoration is gauged. Furthermore, prolonged shifts in climate may produce changes in species distributions, which may mean that restoration targets need to be redefined. Thus, in stream restoration in such situation, “recovery” from drought may become conflated with the responses to restoration. We present a conceptual framework for considering these issues based on data we have collected to evaluate the ecological responses to various stream restoration activities (riparian restoration, flow regime restoration and habitat reconstruction). Dealing with responses to powerful natural disturbances, such as drought, in the context of restoration is a major challenge-a challenge that may become more widespread and increasingly complex as climate change strengthens.