Both Knave and Tunnels and trolls share a couple of properties: crawl games, fairly cartoony encumbrance (though in different ways) and a dicing system with a fairly difficult baseline difficulty.
I want to talk about the task resolution. (Practically every game resolves tasks and conflicts in some way; these ones resolve tasks explicitly and conflicts as a consequence of several discrete task resolution checks, or sometimes just one task resolution check.) Both of the games have a baseline chance of success of around 40 % with a big inaccuracy involved, yet it does not make a difference here. Knave does have a possible advantage die to adjust this into an easier roll, but still, the difficulty remains fairly high.
In Knave it was about orienteering when a range of mountains is clearly visible to give a direction; in Tunnels and trolls it has been about slashing webs with a magically sharpened knife, crossing pits and having a few people hold one up as that one carefully puts weight on a pit trap to open it.
These kinds of tasks become highly volatile; either the game master gives an automatic success or it becomes a highly risky act. There is no way to dice "it usually works, but maybe not this time".
If we think of real life risk-taking, having, say, 1/10 chance of getting lost would be seen as highly risky, and likewise 1/10 chance of falling to an unknown destiny. Rpg design often aims towards somewhere between 1/3 and 2/3 for the success chances, plus, especially in more modern games or "story games", significant failure stakes.
This creates a very particular, heightened or dramatic, kind of narrative, or in case of crawl, a game of taking risks rather than managing and minimizing them. Or maybe another formulation: the friction of war is either removed or exaggerated into a dramatic obstacle. This is very noticable after having played tens of sessions where 1/6 (or 5/6) chance with several bonus or penalty dice are rolled as a normal part of the procedure; or sure, my monk is using their psionics with roll under of d20 (20 explodes) against 21, and there 42 would be a second degree failure with possibly significant consequences, and 63 third degree failure, etc, so these have a chance of less then 1/400 and 1/8000, respectively. Yet still we roll, because psionics is dangerous and these might happen. I always have to make the active decision to take the risk to fry the character's brains, even if it is fairly small; exactly how small should the risk be to be worth it? That is for me to decide.
I am interested in further thoughts around very likely and unlikely odds, and especially constructive use of such.