Possession comes to mind. Getting to Possession 1 turn sooner means snowballing where the first there gets an increasing number of turns more than the opponent. E.g. on turn 10, you get 2 turns I get 1. On T11 you get 5 turns, I get 2.
Similarly I would say on a lot of boards, gaining Outpost at the right time from the Bm is just utterly broken. For high level play, most Bm boards have a viable Oupost engine (singleton enablers include Guide, Den of sin, Scheme, Wharf, Enchantress for instance). Once you both have reliable enough engines they get literally double the opportunities you do. There are so very few things can possibly hope to come back from that. Sometimes you can hope for a three pile, sometimes you get an enabler for a Megaturn, sometimes you get so crazy valuable card (e.g. Tournament) ... but facing half as many turns is just insanely overpowered.
Other setups include things like the opponent have two Knights in play to your zero with only one gain per turn with the other 8 Knights in the trash. You might be able to eke out a win here if you are close enough to already winning, but when your deck is going to eventually be decimated you need a VERY strong lead not to inevitably fall behind.
A last one is the opponent building a stronger Golden deck than you. Winning the Fort split 6/4 or 7/3 with Bishop out makes it very hard to compete if there are no other villages. Good luck piling out before they wrack up enough VP to be out of your league. Similar things can happen with Gardener/Amb (particularly with any kingdom green) or some Tomb setups (like winning the village split to run 2 Beggars/3 Chapels when your opponent cannot). Again you need to be far enough from winning that you cannot possibly force game end before they will inevitably surpass you in points.
None of these are all that common, but they all represent states of the game where you are both behind and falling further behind each turn. The sorts of terrible draws your opponent would need to give you a chance are exceedingly far out on the bell curve (e.g. >3 sigma). Worse, many of these setups require very little skill to play optimally so chances of opponent mistakes are very low.
Thankfully all of this sort of thing tends to be rare.