Torment feels to me like a design concession that players can get vastly more powerful than originally planned, their defensive stats or strategies effectively blanking the damage dealt by most monsters in extended. Good luck being a speed 10 melee monster against casters spamming L9 spells at max distance, or even against melee dudes with plenty of AC to tank your hits who will kill you in 1-3 hits at mindelay. Even monsters that attack at range will often find themselves incapable of doing any significant damage to melee characters, and incapable of taking enough turns to deal damage against casters (either because of a high damage spell, popcorn allies getting in the way, or popcorn summons by the caster). Smite targeters are a little better off, but their damage rolls are usually not high enough to force the player to reposition. The only "winners" in extended are the proactive monsters who make it a nightmare to find yourself in LOS with, either because of the threat of immediate death or the situation turning much worse. Torment is of course the shining example, but hellfire, paralyze (and malmutate maybe? idk if it's less bad now) are also up there. Whereas most extended characters can get smited or melee'd or shot at range for several turns in a row (which is long enough usually for them to dispatch everything), the same can't be said if one or more torment-knowers decide to torment instead of walk harmlessly forward on any given turn. It's such a powerful option, making it so terrible for the player to be in a group situation against such an enemy that they must either 1) run away to kill it safely alone, 2) nuke it as fast as possible, or 3) get one of the rare effects which mitigate it. Since tormenting monsters can torment the very first turn they enter LOS, and trying to nuke every torment source will inevitably lead to you eating some torments since you can't reliably one-shot anything in the game, most people choose option 3 to avoid getting cheesed out as best they can, or maybe (if you'll allow the editorialization) because they don't find torment any fun to play around normally.
Now, yes, that was all to say that torment is a "winner" in extended, but I want to go a little further to make the case that torment is not just powerful but overcentralizing. I'll start by talking about other "winners" I listed, as they are all egregious but for one reason or another not overcentralizing.
- Hellfire - this is a strange one because the only way to resist is is Max HP/damage shaving. As such there isn't really a way to build "against" it so it doesn't inform any strategic decisions for the player. Tactically it is about as scary as torment, maybe just a little less bad since 1/2-to-2/3'ing your HP is probably more dangerous than a powerful die roll. It is also less common than torment afaik.
- Paralyze - The only way to resist is high MR of course, which most extended builds will have (XL 27 helps). I suppose fearing para or other bad enchantments can warp a build to use an MR+ item or follow Dith or Trog, but I think that's basically unheard of because para is so rare, has to be chosen from among a spellbook of otherwise-useless spells, and beam-targets so it's hard to get hit by it when it's most dangerous (lots of monsters around). Floating eyes work differently and they were the main thing I was thinking about actually (where you never want to be in LOS with them) but once again like Hellfire there isn't really a strategic way to play around them. AC, max HP, and good positioning are also secret resistances to Para (well, I mean resistances to death from para). Also note that nowhere am I contending para (or any of these other mechanics) are good, only that they don't warp player decisions extremely. I think para is terrible design but that's a topic for another day.
- Malmutate - Certainly annoying and it makes me play around malmutators in the same kind of way as tormentors, but again this is beam-targeted and somewhat rare (though maybe it's actually just on par with torment for rarity). The only reliable way to resist it is Zin, but I think with !mut changes and abyss farming if you need it it's not ever a lethal thing so it doesn't really warp gameplay enough to make players feel like they're missing out on something big by not using Zin.
So back to torment, where does that leave us? Like these other extended game "winners" it is hard to avoid, with rN being the frontrunner that only somewhat reduces its efficacy. rN+++ is still unquestionably a must especially in Tomb, but getting randomly double- or triple- tormented will still leave you hurting (down to 30% of initial after 3 with full rN). You can follow Kiku but this is basically just rN+++ for free, which at least does not restrict equipment slots as much. You can learn necro, or try to live with it, but it will be the largest source of damage and the biggest threat throughout the game for sure if you do this. Note that unlike with the other three "winners" I mentioned it isn't really viable to not pick one of these. Deciding "I will play a character who has no rN and takes 50% HP against torment in extended" is not really an option. Your hand is forced in one direction or another. Sacrifice equipment to get rN+++, sacrifice god to get Kiku, sacrifice skill investment into other places to get Necromutation. Did Ru design this mechanic?
Note that the barring lich form, torment is still a must-deal-with above almost anything else. Thus
even with proper resistances it is hard to 'beat' torment. This is why I view it as overcentralizing. It restricts choices in endgame builds and is still powerful even when steps are taken against it.
So, is having an overcentralizing mechanic like torment so bad? True, it more or less mandates resistances and is still one of the most lethal threats to players anyway, and true that it is quite prevalent and cannot be avoided or mitigated by any other defensive stat. But what should the endgame look like, if not a wasteland of popcorn and a few high power threats like the rest of the game? Well, my main contention is that endgame popcorn is even worse than early game popcorn proportionally, and it could be buffed to be more respectable. At the same time torment and other "winner" mechanics could be toned down to balance this. Make enemies that do good damage but can't live long enough beefier, and make enemies that can never do any damage to endgame player stronger or give them some tactically relevant ability (like Mesm or something new). Additionally do not spawn enemies that are far too weak for endgame just because they are thematically appropriate (such as low level demons).
If you think about how torment plays out in non-lich-form builds, it roughly doubles the damage output of every other enemy in the fight until the player can reset. Due to it working off current HP it's hard to not have more than half HP in a fight with a tormentor, often less if you got unlucky. Only in these straits are most extended enemies much of a bother. Torment is also super game-able, separate the torment-user from the pack and they become harmless again. Therefore weakening torment and strengthening the default output of enemies in extended would make it less overcentralizing and have the benefit of being a little harder to game - well, you can still pick off packs safely no matter what I suppose, it just becomes less painfully obvious that it is the best choice with this change.
I think some monsters like Antaeus and Ancient Liches hit hard, so giving other monsters a shred of whatever kind of power they have might be enough. Orbs of Fire are also a good example, and they show that the game isn't completely void of endgame level threats that don't just LOS target you and halve your HP. I want some more dangerous enemies in that vein, a kind of middle ground between torment and popcorn.