In a previous study, a culture of the electrical-current-producing bacterium Geobacter sulfurreducans str. DL1 was evolved to select an isolate with an ability to produce more current. However, the isolate (termed KN400) was shown to contain more than 27,000 SNPs and 139 unique ORFs, indicating that it could not possibly have evolved from G. sulfurreducans str. DL1. It was therefore assumed to be an external contaminant that was inadvertently introduced during the experiment. By sequencing a gene that differed by 14 bp between the DL1 and KN400 strains, Shrestha et al. now show that the KN400 strain was in fact present in very low abundance in the initial G. sulfurreducans str. DL1 culture, despite the DL1 strain having undergone the standard approach of serial dilution and repeated re-streaking of isolated colonies on solid agar to ensure its purity. Furthermore, previous deep sequencing of the purified DL1 strain to 80-fold coverage failed to detect KN400-specific sequences. These data emphasize the ability of extremely rare variants to go undetected by deep sequencing and the inadequacy of the repeated-streaking approach to obtain a pure culture. The authors propose that to avoid undetectable contamination, cultivation from a single cell is required, which could prove technically challenging for many environmental microorganisms.