That's not why I'm doing this. I put up the forum post hoping that the community would notice something Jeremy and I missed.
It's fairly clear to me that you've put a lot more thought and [research? personal experience?] into the fine points of this mechanic than the average forum reader, so I'm not sure if you're going to gain much defending your ideas against the superficial criticisms we can provide.
And there's no personal experience, but my ex's family suffers from stress-triggered dissociation. My ex's little sister especially tends to see her dead father when she's stressed out.
The issue here is that I can't account for how deep in character the player is. If they're deep in character a few hints of failing sanity will push them into the realm of madness. (As they, for example, have their character rush up the stairs into their attic with a gun, searching for the source of an imaginary creaking sound.) If they don't get into character at all, there is no amount of effort on my part that will get them to behave like a crazy person. This mechanic is really just some non-standard encounters and interface screws. It relies on player immersion to be any more than that.
Nevertheless, here is my superficial criticism: it seems like you haven't made a clear distinction between the "real" mental health of the avatar and the perceived mental health of the player.
If I have a lot of concrete statistics to manage and specific rules and interactions to consider, along the lines of a pen and paper rpg, then I expect to be loosely bound to my avatar. I get plenty of information (more than he would reasonably have access to, e.g. his precise level of healthiness) but my means of acting on that information are mediated by the rules of the game and his current state. A crazy avatar, in this sort of game, might do things I don't expect.
On the other hand, if I am forced to perceive things exactly as the avatar does, then I would expect to be more tightly bound to my avatar (there should be relatively few cases where I don't have control) but at the cost of my unrealistic introspection with respect to statistics. In this sort of game, I would not expect my avatar to go crazy, so much as the game to try to convince me that I was crazy.
Both could be enjoyable games. In the past, I've played survival games that hewed towards either side (not with respect to mental health, just in general), and enjoyed them. But a sort of middle ground option might easily get confusing.
As for the amount of information, all information in this game is skill-dependent. (For instance, medicine determines health information.) The speech skill determines how much information you have on your mental health and emotions. Only at 100 speech do you actually get a number. The rest of the time you rely on descriptive words.