Constraint Based Methods had been successfully used to simulate genome-scale metabolic behaviors over a range of experimental conditions. In most applications, environmental constraints are parameterized, and the use of metabolic reactions and corresponding genes is the direct consequence of the tuning of these parameters. However, in evolutionary studies, the problem is different: one knows the relative importance of reactions and one seeks environmental conditions that could explain such a biological fitness. This study details this modeling paradigm change and discuss a putative formalization of such a biological problem in the form of a Mixed Integer Bi-level Linear Problem (MIBLP). Unfortunately, solving a MIBLP is difficult, paving the way for the need of further constraint based method developments for understanding evolutionary processes.