Intensive nutrient removal from wastewater in anaerobic–aerobic systems using granular sludge should rely on optimal balances at biofilm and microbial ecology levels. This study targets the impacts of reactor characteristics and fluctuations in operation conditions on nutrient removal and bacterial community structures by means of microbial and numerical ecology methods. The dynamics of both predominant and accompanying populations were investigated with high resolution on temporal and phylogenetic scales in two reactors operated during 5 months with synthetic wastewater. Multivariate analyses highlighted significant correlations from process to microbial scales in the first reactor, whereas nitrification and phosphorus removal might have been affected by oxygen mass transfer limitations with no impact at population level in the second system. The bacterial community continuum of the first reactor was composed of two major antagonistic Accumulibacter-Nitrosomonas-Nitrospira and Competibacter-Cytophaga-Intrasporangiaceae clusters that prevailed under conditions leading to efficient P- (> 95%) and N-removal (> 65%) and altered P- (< 90%) and N-removal (< 60%), respectively. A third cluster independent of performances was dominated by Xanthomonadaceae affiliates that were on average more abundant at 25 °C (31 ± 5%) than at 20 °C (22 ± 4%). Starting from the physiological traits of the numerous phylotypes identified, a conceptual model is proposed as a base for functional analysis in the granular sludge microbiome and for future investigations with complex real wastewater.