Introduction: Organic waste co-digestion is an attractive option for territories to valorize different types of locally produced waste and optimize biogas production in digestion plants. We studied the dry batch co-digestion of reconstituted racehorse manure (MA) and the organic fraction of residual household waste (HW), in pilot reactors. We identified optimal proportions of co-substrates and evaluated the effects of their initial proportions on microbial community dynamics. Methodology: Batch pilots for dry anaerobic digestion (60 L each) were filled with HW:MA in various proportions (w:w), ranging from 0:100 to 92:8, for a total of 14 incubations. The co-substrates were fully immersed in digested mixed urban wastewater sludge serving both as inoculum and liquid input. The pilots were operated at ~36°C during up to 5 weeks. The conversion dynamics was characterized by monitoring relevant physico-chemical parameters: biogas production and composition, pH, volatile fatty acids (VFA) and total alkalinity content. The composition of archaeal and bacterial communities was determined by 16S rDNA metabarcoding targeting the V4-V5 region of archaea and bacteria, using an Ion Torrent PGM sequencer. Principal component analysis was performed with R packages such as FactoMineR. Results: When co-digesting HW, a transient accumulation of VFA was observed. Maximal VFA concentrations were all the higher as the initial proportion of HW was high and reached up to ~17 g/L. Moreover, a partial inhibition of methane production occurred in the presence of HW, mostly during the 8-15 first days of incubation, especially for proportions 79:21 and 92:8. The total cumulated productions of methane were similar for HW proportions up to 79%, suggesting the adaptation of microbial communities to high VFA conditions. Consistently, clear changes in the microbial composition occurred according to the HW proportion. In the absence of HW, bacteria were dominated by classes Bacteroidia and Spirochaetia; for high HW proportions, Spirochaetia proportion decreased at the benefit of Clostridia and Bacilli. For archaea, Bathyarchaeia members were the most abundant in the absence of HW, while Methanobrevibacter and Methanosarcina genera were selected at the highest HW proportions. Moreover, the proportions of archaea were significantly lower as soon as HW was employed. Conclusions: The adaptation of microbial communities over time to high VFA concentrations suggests the possibility to employ a pre-adapted inoculum, to avoid the transient inhibition phase while digesting significant proportions of HW. Moreover, it opens the perspective of identifying inhibition biomarkers.