The explanatory gap you're seeing is the difference between equilibrium and dynamic analysis. As time goes to infinity, the probability of an organism retaining

The explanatory gap you're seeing is the difference between equilibrium and dynamic analysis. As time goes to infinity, the probability of an organism retaining genes not useful to it declines to zero. Given enough time, that resistance gene, if not directly beneficial by side effect, will be pruned. But evolution takes time. If the resistance gene is on a stable plasmid, if the plasmid confers other fitness advantages, and if the resistance gene itself only infinitesimally reduced fitness, then evolution can take a very long time indeed to reconfigure that plasmid and optimize away the resistance gene. Both camps are right here.