Understanding complex responses of metabolic processes in biochemical systems requires the quantitative description of the dynamic interplay between metabolite levels, enzyme levels, and reaction rates. Generation and use of such models is hindered by the intrinsic nonlinearities of the enzymatic rate expressions, and the uncertainties at the level of concentrations and kinetic parameters. These issues become even more challenging as we consider genome- scale models. Here, we use the ORACLE (Optimization and Risk Analysis of Complex Living Entities) framework, which is an efficient and scalable methodology for the generation of populations of large-scale, non-linear models of metabolism. With this approach, we can explore the properties of a system of biochemical reactions, their dynamic responses, and their potential to maintain a steady state upon perturbations. We can further analyze the effect of integrating data from thermodynamics, available omics, and kinetic data on these properties. To demonstrate the utility and performance of this methodology we constructed a population of large-scale dynamical models of optimally grown E. coli. We used these models to: (i) perform a modal analysis and characterize models associated with biologically relevant time scales, (ii) characterize basins of attraction around the identified steady states. The aforementioned analyses provide valuable insights for the design of synthetic biology and metabolic engineering strategies towards the development of robust whole-cell biocatalysts for bio-sustainable economies.