Deoxygenation in the global ocean is predicted to induce ecosystem-wide changes. Analysis of multidecadal oxygen time-series projects the northeast Pacific to be a current and future hot spot of oxygen loss. However, the response of marine communities to deoxygenation is unresolved due to the lack of applicable data on component species. We repeated the same benthic transect (n = 10, between 45 and 190 m depths) over 8 years in a seasonally hypoxic fjord using remotely operated vehicles equipped with oxygen sensors to establish the lower oxygen levels at which 26 common epibenthic species can occur in the wild. By timing our surveys to shoaling hypoxia events, we show that fish and crustacean populations persist even in severe hypoxia (<0.5 mL L−1) with no mortality effects but that migration of mobile species occurs. Consequently, the immediate response to hypoxia expansion is the collapse of community structure; normally partitioned distributions of resident species coalesced and localized densities increased. After oxygen renewal and formation of steep oxygen gradients, former ranges re-established. High frequency data from the nearby VENUS subsea observatory show the average oxygen level at our site declined by ~0.05 mL L−1 year−1 over the period of our study. The increased annual duration of the hypoxic (<1.4 mL L−1) and severely hypoxic periods appears to reflect the oxygen dynamics demonstrated in offshore source waters and the adjacent Strait of Georgia. Should the current trajectory of oxygen loss continue, community homogenization and reduced suitable habitat may become the dominant state of epibenthic systems in the northeast Pacific. In situ oxygen occurrences were not congruent with lethal and sublethal hypoxia thresholds calculated across the literature for major taxonomic groups indicating that research biases toward laboratory studies on Atlantic species are not globally applicable. Region-specific hypoxia thresholds are necessary to predict future impacts of deoxygenation on marine biodiversity.