Gabapentin is a weird ass medication, and really messes with your mind. I was prescribed it for a few months... You would have to take well over 1200mg to have it actually promote sleep, and that is neurotoxic range... Never mind that neurontin is an iatrogenic/teratogenic substance.
I think my previous remarks about both individual differences and demonisation have been confirmed. 
Here's another perspective:
Therapie. 2010 Jan-Feb;65(1):57-60. Epub 2010 Mar 8.
[Therapeutic drug monitoring of gabapentin]
[Article in French]
Tribut O, Bentué-Ferrer D, Verdier MC; le groupe Suivi Thérapeutique Pharmacologique de la Société Française de Pharmacologie et de Thérapeutique.
Laboratoire de Pharmacologie Biologique, CHU Pontchaillou, Rennes, France. olivier.tribut@chu-rennes.fr
Abstract
Gabapentin is a structural analogue of GABA used in the treatment of the partial epilepsies of adult and child of more than 12 years, in monotherapy or in association with other anticonvulsant drugs. In association, gabapentin presents the advantage of not interfering with the other anticonvulsant drugs. The interindividual pharmacokinetic variability and the saturable absorption are, with the adaptation in case of renal insufficiency, the only arguments in favor of TDM. During clinical studies, the plasma concentrations of gabapentin were generally included between 2 and 20 mg/L. For this molecule, the level of proof of the interest of therapeutic drug monitoring was estimated in: possibly useful.
PMID: 20205997 [PubMed - indexed for MEDLINE]
There doesn't seem to be too much talk about toxicity in practical, in vivo cases.
There is some mysterious allure GABA -ergic substances have, but messing with LTP is a good recipe for amnesia.
Temporary or permanent?
Depends on who you ask. There is very little research done into long-term effects of any psychotropic medications except typical antipsychotics, and even that data focuses on epidemiology, and not fine grain memory issues. The only long-term benzodiazepine withdrawal study I've seen indicates that on average memory problems do not resolve 7 years out. For obvious control reasons the quality of the data is dubious. Longitudinal anticonvulsant studies only look at long term effects of pre-natal exposure, which is likely shows the most extreme cases, and those suggest that most anticonvulsants significantly degrade cognitive function, with the possible exception of lamotrigine, the reason that it's only possibly excluded is because these studies don't include positive controls, most likely because the data would be more damning.
My suspicion is that people having varying degrees of resistance to brain insult, and varying regenerative efficacy. Also, there is likely a great variation in how patients deal with withdrawal, and what healing modalities they pursue. Personally I wouldn't pay much attention to that benzo study based on that last reason alone: the authors themselves suggest innacuracy when they write that selection bias likely negatively impacted the outcome (aka only people who suffered long term side effects responded). What the study does tell us is that for the population that experiences long term effects these effects can be "permanent", and I quote that because likely these people have not exhausted all options to improve the deficit, and also it is notoriously difficult to predict how a patient will recover (in tbi cases at least) even that long out. There is no evidence that the kind of "brain damage" you experience from medication use, barring long term anti-psychotic use, causes the kind of obvious, macroscopic damage, especially scarring, that insults like electroshock or acute head injury are confounded by.
On another note, in another study I read iatrogenic epigenetic "permanent" changes were noted in a wide swarth of sampled psychotropic medications, everything from ssri's to antipsychotics. I can dig up the list, but I read this study last year and cannot recall all medications, but fluoxetine and some anticonvulsant and antipsychotics were tested. The authors concluded that these "permanent" expression changes were possibly the cause of prolonged withdrawal effects. It is however, quite clear that gene expression is modified by many actions, anything from meditation and exercise to drug ingestion, so this doesn't mean much by itself. It would be much scarier if these drugs caused iatrogenic changes, and were unique in their ability to influence gene expression.
As far as gabapentin's neurotoxicity profile, I would dig deeper. First of all, it doesn't need to be overtly neurotoxic to be damaging to mental function. Signs of neurotoxicity, like nystagmus or fever, are only observed when direct cell death occurs. Gabapentin, on a mg^-1 basis is not the most toxic substance, but it absolutely dangerous, and there are high quality studies showing this. Check science direct or ncbi.nlm.nih.gov
