Coastal hypoxia is increasing worldwide in response to human‐caused changes in global climate and biogeochemical cycles. In this paper, we view anthropogenic trends in coastal hypoxia through the lens of disturbance ecology and complexity theory. Complexity theory provides a framework for describing how estuaries and other coastal aquatic ecosystems respond to hypoxia by understanding feedback loops. Can it also be valuable in understanding how these ecosystems behave under shifting (i.e., unnatural) disturbance regimes? When viewed as a disturbance regime, shifts in the spatial (areal extent and fragmentation) and temporal (frequency and duration of events) characteristics of coastal hypoxia can be used to track changes into a non‐stationary future. Here, we consider options for increasing the resilience of coastal aquatic ecosystems to future, unnatural hypoxic regimes. To start, we define desirable states as ecosystems with long trophic chains and slow nutrient and carbon dynamics that produce many ecosystem services. We then work backward to describe circumstances dominated by positive feedbacks that can lead ecosystems toward an undesirable state (i.e., depauperate communities and chemically reduced sediments). Processes of degradation and recovery can be understood in the context of island biogeography whereby species diversity in habitats fragmented by hypoxia is determined by the balance between rapid local extinction, slow recolonization from the edges of hypoxic patches, and opportunities for ecological succession during between disturbance events. We review potential future changes associated with changing global climate and highlight ways to enhance coastal resilience. In addition to efforts to slow climate change, potential interventions include reduced nutrient and carbon loadings from rivers, restoration of aquatic vegetation, and managing for key species, including those that promote sediment oxygenation, that enhance water clarity, or that promote grazing on epiphytic algae through top‐down control.