Microbial communities provide human society with a variety of benefits including bioremediation. To optimize these benefits, it is necessary not only to maximize process efficiency, but also to stabilize the communities over ecological and/or evolutionary time scales. Indeed, degrading a toxic compound through the production of an enzyme is regarded as a public goods game, where the invasion of cheaters (non-producers) is predicted. In this study, we constructed an evolutionary game theoretical model, where a microbial population degrades a toxic compound, and used the model to analyze (i) whether cooperators (enzyme producers) can be maintained, and (ii) how to optimize detoxification efficiency by manipulating the toxic compound concentration and/or the dilution rate. Although cooperators go extinct when cheaters that are equally resistant to the toxin appear, we identified conditions where cooperators with the different level of resistance can invade the population of cheaters and exclude them. However, in the long term, cheaters of the same resistance level are bound to reinvade. To overcome this inevitable outcome, we show that cooperators can be periodically reintroduced to the population to maximize the efficiency and stability of detoxification.