put that beta-lactam back where you found it
One of the common Just-so stories about antibiotic resistance is that bacteria will generate it in the presence of an antibiotic, and then lose it when placed in an antibiotic-free environment. It’s a nice story, and really convenient when one considers all the antibiotics shoveled into food animals or used as prophylactics–but it’s not a story consistent with human experience. If M. tuberculosis et al reverted to antibiotic-sensitivity away from penicillin, we wouldn’t need the whole host of semi-synthetic beta-lactams we’ve got.
I think the problem comes from some implicit assumptions about how the resistance come about, and maybe misconceptions about how populations change under selection. A gene conferring antibiotic resistance isn’t necessarily a de novo formation; it may well be a modification of an existing gene, and it may or may not compromise the function of that gene. Either way, there’s probably some cost associated with the expression of antibiotic resistance, but the survival that gene confers in an antibiotic-saturated environment renders that cost irrelevant.
Organisms adapt to their genetic as well as their physical environs, and in a context in which antibiotic resistance is necessary, that culture of M. leprae (or whatever) will become both highly optimised and resistant. Optimisations in the genetic context of resistance are a different set of optimisations than those outside, making each a separate adaptive peak. Removing the antibiotic will then leave the resistant strain at a relative, but not absolute, fitness peak.
And that is why you’d bloody well better finish your pills–because asexual organisms don’t drift much.