Soil respiration represents the second largest CO2 flux from terrestrial ecosystems to the atmosphere, and a small rise could significantly contribute to further increase in atmospheric CO2. Unfortunately, the extent of this effect cannot be quantified reliably, and the outcomes of experiments designed to study soil respiration remain notoriously unpredictable. In this context, the mathematical simulations described in this article suggest that assumptions of linearity and presumed irrelevance of micro-scale heterogeneity, commonly made in quantitative models of microbial growth in subsurface environments and used in carbon stock models, do not appear warranted. Results indicate that microbial growth is non-linear and, at given average nutrient concentrations, strongly dependent on the microscale distribution of both nutrients and microbes. These observations have far-reaching consequences, in terms of both experiments and theory. They indicate that traditional, macroscopic soil measurements are inadequate to predict microbial responses, in particular to rising temperature conditions, and that an explicit account is required of microscale heterogeneity. Furthermore, models should evolve beyond traditional, but overly simplistic, assumptions of linearity of microbial responses to bulk nutrient concentrations. The development of a new generation of models along these lines, and in particular incorporating upscaled information about microscale processes, will undoubtedly be challenging, but appears to be key to understanding the extent to which soil carbon mineralization could further accelerate climate change.