The simulation of supersonic combustion requires finite-rate chemistry because chemical and fluid mechanical time scales may be of the same order of magnitude. The size of the chosen reaction mechanism (number of species and reactions involved) has a strong influence on the computational time and thus should be chosen carefully. This paper investigates several hydrogen/air reaction mechanisms frequently used in supersonic combustion. It is shown that at low flight Mach numbers of a supersonic combustion ramjet (scramjet), some kinetic schemes can cause highly erroneous results. Moreover, extremely fine computational grids are required in the lift-off region of supersonic flames to obtain grid-independent solutions. The fully turbulent Mach 2 combustion experiment of Cheng et al. (Comb. Flame 1994; 99: 157–173) is chosen to investigate the influences of different reaction mechanisms, grid spacing, and inflow conditions (contaminations caused by precombustion). A detailed analysis of the experiment will be given and errors of previous simulations are identified. Thus, the paper provides important information for an accurate simulation of the Cheng et al. experiment. The importance of this experiment results from the fact that it is the only supersonic combustion test case where temperature and species fluctuations have been measured simultaneously. Such data are needed for the validation of probability density function methods. Copyright © 2009 John Wiley & Sons, Ltd.