But you're doing that in a way such that they're learning a lesson which is untrue. You want to establish the enemy as a threat, so you make this one monster much scarier than it would otherwise be (through use of a meta-game resource); their takeaway is that this is a very strong monster which needs to be dealt with using one of its prescribed weaknesses.
But if they encounter a group of ten such creatures, at a later point, they're going to think that these are all very scary monsters, and they'll need fire or a called shot if they want any chance of putting one down. They'll know through first-hand experience that regular attacks are ineffective. If one of them goes down from a shot to the torso - if anyone even attempts that, given their certain knowledge that it would be ineffective - then it creates a plot hole.
Or I'm playing up narrative elements over a slavish devotion to realism - the first time such a foe is encountered, it's a terror. The next time, the heroes have figured out how to deal with those foes, and they're weaker by comparison. I'm running something that resembles a story, with a particular pace and particular narrative beats, albeit guided by player choices, rather than a simulated chain of events to which I'm just an impartial judge (which seems to be a sin to some people).
I dislike the notion that an NPC must always exist in a specific, singular state (indeed, that element has more in common with computer games than anything else - NPCs as a pre-set and immutable package of stats). Fights aren't fair, they're not all equal, and nobody should enter battle with the certainty that "I'm fighting X, so I'm probably going to win", and the uncertainty and variation that comes from an additional factor that they can't 100% plan for is a valuable one for keeping action scenes tense, and keeping foes as viable threats regardless of how powerful the PCs have gotten.
I guess what I'm getting at is this:
If you've already established that you're not playing an antagonistic GM (in the vein of Basic D&D), then you've decided that you're not going to kill the PCs unless they're really asking for it (either through stupid or heroic actions, where death would be suitable to the genre). You just want to scare the players, and make the story more dramatic. At that point, why do you need a meta-game resource to regulate that? Why can't you just fudge some dice rolls, whenever you feel like it? How does it improve any aspect of the game when the Big Bad goes down before it can even act, because the GM spent too many Dark Symmetry points on keeping the Dragon up for another round? Is it just a way to keep the GM engaged with the game aspect of it, by giving a resource to manage?
I don't
need a metagame resource. But this isn't about need. I like the outcomes it produces, and the effects I can create with it. I recently did a revision of the Infinity RPG playtest that frames "Heat" (the name for Threat in that system) as the kind of noise and unpleasant attention that gets in the way of covert ops. High Heat means that enemies are more alert, that the situation is risky and filled with peril. In one of the adventures in the Mutant Chronicles line, I used Dark Symmetry points to serve as a stealth and subtlety system, where the total number of Dark Symmetry points represents the security at a high society party being infiltrated, and players contribute Dark Symmetry points to the GM through actions that might draw undue attention. It's a tool, and while it isn't essential, it has certainly been useful, and I enjoy seeing what I and other GMs can do with it.
You're right that it is, in part, a way to keep the GM engaged with the game aspect. I find that a lot of RPGs overlook that the GM is a participant as well, and while that's fine for some, that isn't always ideal for every style of game. I got into game design in part because I like the rules side of RPGs, particularly when the rules interact interestingly or complement the roleplaying side of things. I dislike systems that just "get out of the way" when roleplaying happens, as if RP and G are anathema to one another. I dislike systems that assume that the GM's only real interaction with the rules is pre-session prep and teaching the players, because I want to play too, rather than just serving as Master of Ceremonies for everyone else's fun (having spent two years dealing Roulette and Blackjack professionally, I've done "you're playing, I'm just facilitating play").
Beyond that... some players don't like the idea of a GM fudging the dice rolls, and don't respond well to the idea of it. I've done it with some games, but I prefer systems where I don't
need to, because that kind of flexibility and authorial power is a part of the system, rather than using the GM as a way to patch the system if it creates undesirable outcomes (a good GM doesn't need a system, a bad GM can make even the best system awful).
And, once you get down to it, GMing is a daunting prospect for those who've never done it before, particularly with systems where there aren't guidelines, limits, and structure. GMing where there's a structure in place is an easier prospect... and we want more GMs, because that's the only real way the hobby can expand.
It's also worth remembering that Threat is one part of the system, not the whole of it.
Broadly, test difficulties do not scale linearly, but the maths of the system is obfuscated into relatively straightforward player choices. For a character rolling only two dice (the default, buying no extra dice), with an Attribute of 8, and Expertise 2 (adds +2 to the attribute for that test) and Focus of 2 (any natural roll of 1 or 2 on any d20 counts as two successes) each in a given skill, passing a Difficulty 1 test happens 75% of the time. Difficulty 2 reduces that to 25%. Difficulty 3 reduces it to 5%, and Difficulty 4 to 1%. Difficulty 5 is absolutely impossible to reach on two dice. Hitting the high difficulties reliably means buying dice. Threat is the most obvious way to do that, and it's universally applicable... but it comes at the cost that the GM might be able to do the same (or something equivalent) later. Spending Luck - a limited player resource, difficult to recover - is more reliable and comes without adding fuel to the fire (a Luck point spent gives you a d20 that's already rolled a 1 - no risk of Repercussions, and if you've got Focus, an automatic two successes). Assisting a task lets one character add a d20 tested against his skills to someone else's test. Certain skill tests have specific ways of boosting dice pools - Mutant Chronicles and Infinity let you spend Reloads (an abstract unit of ammunition) to get bonus d20s and damage dice on attacks), often drawing on some limited circumstance or player-side resource. A few talents let a character buy two dice for each Threat on particular tests, giving them a better deal for the cost. There are a few ways to get those extra dice, not only Threat. Threat is the baseline, the universal option.
From that perspective alone, Threat is one of the ways that the player hits high difficulties, but it's otherwise not essential. And it's entirely possible to play that way.
However, there's also Momentum to consider. Getting more successes than the minimum required is advantageous - those successes become Momentum. Momentum is a degrees-of-success mechanic, so the more you get the better you succeed, but it's designed to be more versatile than that - a character could spend Momentum to take extra actions, succeed at a task more quickly, achieve more, give aid to an ally or hinder an enemy, etc, or all of the above, if they spread around what they spend their Momentum on. It serves as our 'stunts and manoeuvres' mechanic in combat, with the added advantage that you can see how you did before you choose what stunt you're doing. More than that, player characters can save spare Momentum, passing it into a group pool that any PC can draw from, so one character's successes can benefit someone else. Consequently, succeeding really well on a test is something to aspire to a lot of the time, which makes buying those extra dice more temping.
At that point, it's entirely possible for a single player to pump Threat into the pool, then pass on the benefits to his allies (who aren't buying extra d20s) as Momentum.
