Nitrogen contamination of water sources poses significant environmental and health risks. The sulfur-driven simultaneous nitrification and autotrophic denitrification (SNAD) process offers a cost-effective solution, as it operates in a single reactor, requires no organic carbon addition, and produces minimal sludge. However, this process remains underexplored, with microbial population dynamics, their interactions, and their implications for process efficiency not yet fully understood. To address this gap, this study analyzed microbial populations in a 0.8 L fluidized bed reactor performing sulfur-driven SNAD under increasing nitrogen loading rates (NLR), ranging from 11 to 105 g N/m3 d. The process achieved 93.5 % total nitrogen and 95.1 % ammonium removal at a hydraulic residence time (HRT) of 1.8 days. However, when the HRT was reduced to 0.96 days, nitrate removal instability occurred, reducing the nitrate removal efficiency to 42 %. Although increasing the HRT improved performance, two additional instability events were observed in subsequent stages at HRTs of 1.2 and 1.03 days, where nitrate removal efficiencies dropped to 11 % and 39 %, respectively. Functional analysis showed that NLR negatively impacted the proportion of sulfur-oxidizing bacteria, which was correlated with high nitrate levels in the effluent, although ammonium oxidation remained stable. Ecological network analysis revealed positive interactions between ammonia-oxidizing and heterotrophic bacteria, supporting nitrification stability. However, it also uncovered negative interactions between heterotrophic bacteria and sulfur-oxidizing denitrifiers, such as Dyella and Thiobacillus, suggesting these negative interactions contributed to temporary nitrogen removal problems in the system. This study highlights the importance of functional microbial and ecological network analyses over traditional metataxonomic approaches in understanding SNAD processes.