Low lethality is how gamedevfags (attempt to) mitigate disastrous combination of their players' butthurt and their own inability to brain.
Generally, there are 4 main types of attacks:
- Decisive - achieves a tactical goal/persistently deprives enemy of a capability. If I kill an enemy wizard so they can no longer fireball, or blind enemy archer so they can't shoot, disarm a guy or put an arrow in enemy's warrior knee so they can no longer mover around and attack/defend/take cover I have performed a decisive attack.
- Forcing - forces enemy to do something I want them to, either directly or by threatening with nasty consequences. If I dunk a grenade into enemy's cover to flush them out (or start casting powerful AoE spell for similar effect), or spray that cover with suppressive fire to force them to keep their heads down and allow my dudes to close in unmolested I am performing a forcing attack. Same if I inflict cancellable damaging effect (for example set them on fire). The consequences need to be painful.
- Attrition - doesn't achieve anything interesting by itself but deprives enemy of some amount of interchangeable resource. If I sneak up on superior enemy force and kill or hobble some of their pack animals carrying their supplies or set their supply cart on fire I have performed attrition attack. Sufficient attrition usually amounts to one of the previous attack types.
- Gamble - can potentially achieve something game changing but with low enough probability that you shouldn't count on it. For example spraying enemy with ranged attacks from way beyond effective range (hey, might be a crit...) or casting your typical save or die spell. If it is successful it can amount to one of the other
Of course, those types aren't clearly delineated and depend on the scope and time frame (if I poison someone forcing enemy to spend healing supply or face consequences was it attrition or forcing?), but they should be useful enough.
Now the problem is that gamedevfags concentrate on the least interesting attack types (the last two) that make for the worst possible gameplay almost exclusively.
The problem with gamble is obvious - that it is, well, a gamble. It depends on luck. Involving trace amount of luck makes for more interesting and tense game just like adding some salt to your dinner tends to make it taste better, but try eating just a large pile of salt.
The problem with attrition is mostly the opposite - it's an obvious filler, plain and boring - think just plain potato puree. It can be an useful buffer layer on top of interesting stuff but it's never interesting on its own, since it basically amounts to putting progress bars on stuff.
If attrition and gamble are the only tools in your box you physically
can't make interesting gameplay. No matter in what proportion you mix you plain mashed potatoes and salt you won't end up with particularly exciting dinner.
The question of low vs high lethality
as posed in the context of typical RPG gameplay is falacious - the only winning move is not to play.
What good gameplay actually needs is mostly a combination of decisive attacks and forcing attacks, with pure decisive attacks needing some specific conditions.
Some attrition might be inevitable when tracking resources, same as a little bit of gamble, but the role of both should be subservient to almost deterministic decisive and forcing mechanics.
Consider a character attacking another.
In typical RPG it's something between a gamble (roll dice to see if they die) or attrition (HPs get down a bit).
Now, what should happen instead is characters performing actions to set the enemy for a killing blow (or disarming blow, or a knockout, etc.) and enemy attempting to mitigate while doing the same on their side.
Buffers, preferably mixed with nondeterminism only really help by providing time windows for doing stuff - for example by allowing you to gauge how long the shield guy can stand their ground in the doorway.
Now, when I said almost deterministic I meant there should be a measure of uncertainty in there - just enough to throw off "perfect" plans.
The rule of thumb, however should be that, for example, if you give me a clean shot at your noggin for whatever reason, YOU should expect to die. I, however, should be prepared for your survival, unlikely as it might be.
So, the combat should be very lethal IF/WHEN it becomes damaging, but not necessarily very lethal overall, and it should actually be most interesting tactically before it becomes damaging.