Living cells can be modelled by successively imposing known constraints that limit their behaviour, such as mass balances, thermodynamic laws or enzyme capacities. The resulting constraint-based models enclose all the functional states that the modelled cells may exhibit. Then, predictions can be obtained from the models in two main ways: adding experimental data to determine the state of cells at given conditions (MFA) or invoking an assumption of evolved optimal behaviour (FBA). Both MFA and FBA predictions are typically performed at steady state. However, it is easy to take extracellular dynamics into account. This work explores the benefits of using possibility theory to get these dynamic predictions. It will be shown that the possibilistic methods (a) provide rich estimates for time-varying fluxes and metabolite concentrations, (b) account for uncertainty and data scarcity, and (c) give predictions relaxing the optimality assumption of FBA. On the other hand, these methods could serve as basis for monitoring and fault detection systems in industrial bioprocesses.