Background In cystic fibrosis, dynamic and complex communities of microbial pathogens and commensals colonize the lung. Cultured isolates from lung sputum reveal high inter- and intraindividual variability in pathogen strains, sequence variants, and phenotypes; disease progression likely depends on the precise combination of infecting strains. Routine clinical protocols, however, provide limited overview of the colonizer populations. A more comprehensive and precise identification and characterization of infecting strains could therefore assist in making corresponding decisions on treatment. Results Here, we propose a minimally invasive and culture-independent way to monitor microbial lung content. We describe longitudinal tracking of four patients with cystic fibrosis over several months, including repeated sampling and bulk DNA sequencing of sputum. We find that the taxonomic identity of individual lineages can be easily established. Crucially, even superficially clonal pathogens can be subdivided into multiple sub-strains at the sequence level. By tracking individual allelic differences over time, an assembly-free clustering approach allows us to reconstruct multiple strain-specific genomes with clear structural differences. Conclusions We emphasize the advantages of metagenomics sequencing of longitudinal data for taxonomic identification of microbial colonizers in cystic fibrosis lungs. Our study is the first to track sub-strain pathogen dynamics using a culture-independent approach, with a perspective of high-resolution routine characterization of disease progression that may allow to more accurately monitor treatment outcomes and tailor treatment.