Future climatic projections suggest greater and more variable temperature extremes which will have important implications for grassland species dynamics and productivity. Over time grassland community stability responses to disturbances such as climatic stress or anthropogenic eutrophication are often influenced by grassland species diversity. However, when such disturbances co-occur, grassland responses can be unpredictable owing to shifts in competitive interactions between species. Given that temperature can influence soil nutrient availability, understanding how grasslands will respond to temperature stress together with changes in soil nutrient status may provide important insights for grassland management. This study aimed to explore how grassland productivity magnitude, variability, and stability of unfertilised (characterised by high species diversity and dissimilarity, and more even abundance distribution across species) and fertilised (characterised by low species diversity and dissimilarity, and less even abundance distribution across species) grasslands responded to climatic stress both annually and over three year periods. We utilised long-term (approximately 60 years), in situ grassland aboveground rain use efficiencies (RUE) to understand how the temporal stability of RUE and its constituents (temporal mean and standard deviation (SD)) changed across mean maximum temperature and maximum temperature SD. Maximum temperature SD was a better predictor of RUE metrics than mean maximum temperature. Mean RUE and RUE SD both increased as maximum temperature SD increased, but RUE SD increased more rapidly in unfertilised grasslands than fertilised grasslands. Unfertilised grasslands, therefore, became destabilised in response to variation in temperature stress. Greater RUE stability may result from dominance by grass species (which may be more resistant to water-related stress than species of other functional groups). Unfertilised grassland RUE instability could, in this case, be indicative of a community responding to its variable environment. Unfertilised grasslands may, therefore, be more capable of adapting (rather than being pre-adapted) to stress suggesting that unfertilised grasslands may outperform fertilised grasslands when multiple stressors are present.