By "precisely offset", I mean that although Tracking is often considered a card with a benefit of selecting a good card out of 3 with a "downside" of discarding 2 other cards from your deck, the "downside" actually is non-existent unless the game goes to fatigue. I agree that there could exist decks where there are a certain 2 cards you require to win and you intend to go to fatigue to find them, but I can't think of any such decks in the game currently, because they'd be far too unreliable. Miracle Rogue is the closest I can think of but it's really only Leeroy that's a critical win condition in that deck--everything else you can stand to lose 1 of and be fine. (Though even there it is possible to win without Leeroy.) I guess you'd be sad to turn up Leeroy-Shadowstep-Shadowstep from Tracking.
I 100% disagree that it's analogous to Lookout. The Dominion equivalent of going to fatigue in Hearthstone is shuffling, which happens all the time, so the fact that Lookout trashes a card instead of discarding it is a huge difference. I agree with you though that Lookout is underrated because people have a negative emotional reaction to trashing cards from their deck. (Even the fact that it's good to trash Estates and Coppers is a learned behaviour. I think most people don't instinctively want to use Chapel to do that.)
So 2 things I'm saying that I think are maybe getting lost/confused:
1. "precisely offset" to me sounds like there some sort of equivalence of some sort. I think it's way more than offset by the positive aspects, and I think you agree; I just don't understand the terminology.
2. I don't think Tracking is like Lookout, I think the psychological effect is similar. It's doing something really good, but for some strange reason people are scared of it because there's a rare case when it does something bad.
1. Carefully defining "precisely offset" is what I meant by "I haven't gone through the math".
First we need to think about what matters about your draw deck in Hearthstone. You can't influence its order and you can't learn anything about its order, so every ordering is equally likely until you draw from it. But there are certain cards where the contents of your draw deck matter outside of normal drawing (Mad Scientist, Thoughsteal, etc.). I'm going to ignore those cards, although there's a reasonable argument that running Mad Scientist makes it pretty bad to discard traps with Tracking. With that assumption, we only care about the normal draw step at the beginning of your turn.
Imagine a card X with effect "draw 1, discard 2 from the top of your deck". Because every ordering of your draw deck is equally likely, that effect is exactly the same as "draw 1, discard 2 from the bottom of your deck". So, if you don't fatigue, it's clear that card X's discard effect has no effect on your probability of winning the game. That's what I meant by "precisely offset": the downside of discarding 2 off the top is offset by the upside of digging 2 deeper over the course of the game. I think it's more useful to simply not think of the discarding/digging as a downside/upside at all.
The tricky thing is how to adapt the analysis to Tracking. Obviously its effect is strictly better than card X's, because you get the best card out of the three instead of a random card. But then the discarding isn't precisely offset by the digging anymore, because you're discarding 2 of 3 vs digging for 2. Still, that's where my intuition is coming from.
2. Your point is good, but the balance of the randomness is different with Lookout. Lookout falls in the general Dominion randomness theme of "usually something mildly good happens, but occasionally something quite bad". Usually you get to trash a junk card (mildly good), and occasionally you must trash an engine component (quite bad). Granted, in a zero-sum 2-player game, mathematically any randomness like this is flipped when viewed from your opponent's perspective, but it's more visible when it happens "to you". For example, when your opening buys miss the shuffle (occasional, quite bad), you notice, but only eagle-eyed players notice when their opponent's opening buys miss the shuffle.
With Tracking, on the other hand, discarding good cards with it happens just about as often as finding good cards by digging deeper that you otherwise wouldn't have, it's just that the discard is more visible. People are less likely to notice a critical topdeck that they wouldn't have reached without having played a Tracking several turns earlier in the game.
Hearthstone has a finite deck and you lose life if you can't draw. Tracking removes three cards (itself and two others) to draw one more so the downside of emptying your deck is real. Tracking is also useless if played with an almost empty deck.
These statements are true, yet in practice Hunter simply doesn't fatigue often enough for it to be relevant. Admittedly I've never seriously mained Hunter, but I've played it a bit, watched it a bit, and played against it, and in my experience even midrange Hunter typically only fatigues if it got off a couple mega Buzzard-UTH combos, in which case it's probably winning by enough to not care that you ran out of cards.
I'm fairly sure that an all-out control Hunter would not want to run Tracking for the reasons that you point out, but that's not an archetype that anybody is seriously playing at the moment.