If people with poor social skills improve said skills on a drug, their brains remember how to do those things. This is also true for, say, antidepressants; often a depressed person doesn't have much of a life, they take antidepressants, find meaning, and later they don't need them anymore. While it's commonly complained that, say, SSRIs are not "curative", in practice it isn't too rare for someone to take it for a period of time and then quit to find themselves happier; it's just that said period of time may be as long as five to ten years. Tolerance is a competing phenomenon. Another way to say this is that anxiolytics cause people to give themselves
exposure therapy, which eventually resolves the problem.
Unfortunately the persistent benefits seen with some medications do not generalize to all medications, and in particular the
long-term effects of benzodiazepines can be terrible. But this isn't a fault of addictiveness;
the long-term effects of opioid treatment for chronic pain are much less onerous,
as long as the medication is not abused. Instead in this case it seems that benzodiazepines are simply bad for you. In fact it may be the long-term toxicity of benzodiazepines that scares people away from other anxiolytics.
An aside: benzodiazepines essentially "slow down" the whole friggin' brain. GABAa is the most widely-distributed inhibitory receptor in the brain, and most benzos bind nonselectively to GABAa-linked channels. It doesn't surprise me that this frankly crude inhibition of everything leads eventually to smaller brain sizes.
In other news: how easily of a time do you think
the authors of this study had falling asleep the night after they submitted it for publication? e_e