I'll point out that the current two-process model we have for sleep pushes the brain towards waking up after a while; Process C (circadian) on a periodic basis and process S after the brain's need for sleep is satisfied. That would need to be counteracted. I'll also point out that there's some evidence that even a few months of sedation will cause lasting cognitive damage, e.g. https://www.nejm.org/doi/full/10.1056/NEJMoa1301372, yet sedation for years, if not decades, would be necessary. The brain is not a machine; it does not shut down when asleep or sedated.
> And you still shouldn't assume that death is preferable to solitary confinement or sensory deprivation.
It is, of course, a personal decision and it might even be said to be the ultimate personal decision. However, not understanding the implications of one's choices, as best we know them, is unwise, don't you think?
As such, there are no implication to being a brain in a vat other than the fact that you don't experience the outside world. It says nothing about what you do experience. And it certainly does not imply permanent damage in the same way that death does.
Since none of that is likely to happen here in the real world, ever, it makes far more sense to take a hard science approach based on what is known to be at least theoretically possible today.
We already know what happens to prisoners placed in solitary confinement; the effects on mental health are quite damaging over the long term. There have also been studies on partial sensory deprivation [0] and impairment of cognitive abilities and hallucinations began occurring within days. I can't find the appropriate references at the moment but total sensory deprivation, which is closest to the brain-in-a-jar scenario, is even worse than that IIRC.
[0] https://www.motherjones.com/politics/2012/10/donald-o-hebb-e...
> The title of the game/short novel comes to mind: "I Have No Mouth Yet I Must Scream."
Zelazny's "Go Starless in the Night" would be more appropriate and perhaps only slightly less horrifying.