You should check out the Torchlight forum. Since it isn't a competitive game, people seem to think there is no need for balance at all. Who cares if someone has a skill that one-shots every thing on the screen, destroy bosses, works best with only 5 skill points out of a possible 15, and is the skill you start the game with (Outlander Glaive Throw in this case).
It is important to remember that the point of a game is not necessarily to be balanced. Generally the main point of the game is to be fun, though the more e-sports-oriented ones understandably tradeoff some fun for competitive balance. And some non-obvious choices go into making a game as fun as it can be (restrictions that are frustrating sometimes, but without them it's a house without walls, etc).
That said, it's also important to remember that "competitive balance for e-sports purposes" is not the only reason to balance a game.
If in some RPG you have N abilities available to you, and one of them is so underpowered and so uninteresting that there is
never a reason to use it, then your game is literally worse off for having it (interface clutter and confusion, at least), and you should either rebalance it or go to N-1 abilities.
In the extreme case of that, if you have N abilities available to you, and one of them is so overpowered that there is
never a reason to use any other ability, then for each of those other abilities you should either rebalance them or remove them. From a fun perspective, at least; from a business perspective you might just pad the ability count and hope people don't notice until after you have your sales. I don't know of anyone who's done it to the extreme of only 1 worthwhile ability, but I've seen some suspiciously-like-padding stuff.
Same goes for units in an RTS. In a game like AIW there's always units on the low end and high end of usefulness, but if we were just going to leave something on the very low end and not rebalance it then probably it would be better to just remove the unit. That might happen to cleanup drones or whatever at some point, actually, depends on whether we can think of something worthwhile to do with them (the mine-clearing ability could be folded into something else).
Or in other words: if you're going to give the player a choice between N options, make sure there's a reason to pick each of those, or just pare the list down.
That gets a bit trickier when it comes to story-related choices, like I'm sure there are dialog choices in Planescape: Torment that you'd have to be an idiot to choose if your top priority were "win the game", but that's because there's the more meta-level player choice of whether they care about something else more than that.
