A CSTR reactor was operated for 30 days. It was fed with raw cheese whey with an organic loading rate of 30 gCOD/Ld. Hydrogen production varied significantly with a maximum of 0.9 LH2/Ld and decreased after 17 days of operation. The causes of production instability were analysed using different microbiological tools. It was concluded that this decrease was not due to the incapacity to select hydrogen-producing organisms, as shown by the persistence of the Fe-hydrogenase genes in the reactor. Using a molar balance, it was estimated that more than 30% of the acetic production can be due to the homoacetogenesis pathway, although genes from homoacetogenic microorganisms were detected at a very low concentration. The different effects of enhancement and inhibition of hydrogen production by the lactic acid bacteria and their high abundance variation could explain the instability of hydrogen production in this reactor. (C) 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.