While responding to Alex's comments on my review of the ACKS Player's Companion, I was reminded of a position I've held on spell selection for a while now. Alex argued that Dismember and Choking Grip are both awesome spells, and was surprised that my players, despite having access to them, had fallen back to the old standbys of Fireball and Sleep. I discussed this with the rest of the group, and I think we reached something of an agreement on why Dismember in particular didn't see much use.
The primary concerns which motivated our players during combat spell selection were reliability (How often will this spell actually work?), versatility (How often will this spell be useful?), and raw power (When I get to use it and it works, how awesome will this spell's effects be?). A spell which allows a save, requires an attack roll, or has a variable duration is unreliable, while a spell which can only target one opponent, or which has its effectiveness reduced against certain classes of opponents (eg, undead immune to sleep) lacks versatility in terms of the sort of combats where it can be used to its fullest potential. Unreliability and inflexibility can be outweighed if the payoff on success is very high, which is where raw power comes in, but that's a hard sell.
Dismember is pretty unreliable; a save negates all of its effects. It's also single-target, which reduces its versatility somewhat. Worse still is that the single targets you're going to want to burn a 3rd-level spell slot against are probably enemy chieftains and spellcasters, who have relatively good saves. While mortal wounds are scary, the death and dismemberment table introduces another level of unreliability into the spell; while you could sever both of their arms or break their spine, scoring an immediate knock-out, you've got a similar probability of knocking out a few of their teeth or inflicting minor scarring, which is probably not a relevant contribution to most combats. 3d6 damage is nice, but that's on par with fireball successful-save damage to a single target, and only occurs on a failed save. Fireball, by comparison, deals damage (more reliable than negative effects across types of opponent) to a potentially-large number of targets (versatile), and while it gives a save, it's a save for half (more reliable).
Choking Grip is a slightly different case from Dismember. It has the
advantage of preventing an enemy from casting while under its effects,
in addition to dealing damage which under good conditions outstrips that
inflicted by magic missile or other 1st-level damage dealers (though
those conditions are unreliable, and the targets you really want to deal
that much damage to have many HD, hence probably higher saves and more
likely to end the effect sooner). It can also be cast by a 1st-level
wizard henchman, and doesn't scale with caster level. This means that
if you're (say) clearing hexes and have reason to believe the weaselmen
might have a witch-doctor, it absolutely pays to have your low-level
wizard henches learn Choking Grip to lock down superior enemy
firepower. Also useful for winning wizard-duels if you're short on
spell slots, or for killing your master to take his tower and
spellbooks. Arguably useful for preventing dragonbreath, since a choke
closes the windpipe (as opposed to a strangle, which just closes blood
vessels), but good luck getting a dragon to fail those saves for any
length of time. Less than useful for killing massed humanoids than sleep,
though, and that was what we tended to be doing. Also no use at all
against undead, which were another mainstay of our adventuring diet, and
where magic missile saw a fair bit of use instead of sleep.
This is not to say, of course, that Dismember is a bad spell. It is, however, what I classify as a "DM spell", and it is the division of spells into PC and DM spells which is my main thrust here. This category of spells is generally characterized by long-duration negative side effects (which aren't typically useful against monsters due to their generally short life expectancy), single-target nature, and saving throws. Other examples include Blindness, Bestow Curse, and Phantasmal Killer in 3.x (notable for the fact that the target gets two saves to avoid; normally DM Spells aren't save-or-die, but this one is sufficiently unreliable that I've never met a PC who used it, even dedicated illusionists with Spell Focus to boost the DCs). I claim that spells like these are not all that useful against monsters compared to other spells of their level (Web, Wall of Fire, Charm Monster), but they are very scary when you're a PC and they're aimed at you, because you will have to deal with the nasty side effects for quite a while. Thus, while they don't see wide use by PCs, they make excellent spells for NPCs to use against players. Hence, DM spells. Players are often underwhelmed when they select a DM spell for its cool factor, use it a few times, and see their opponents make the saves and suffer no effect, but the DM can go through as many casters and slots as he needs to, and isn't out to murder the PCs to the maximum extent of his abilities anyways. By contrast, players tend to respond well to spells which are reliable and versatile, because they pay a fairly heavy premium for switching out spells and will be confronted with a wide variety of situations with a single set of spells known, and when an unreliable spell fails they've burned an important resource which they may not get a chance to recharge for some time. Hence my criteria for a PC spell, above; it all boils down to risk management and resource management, and DM spells are a risky way to spend the limited resource of repertoire slots.
On the other hand, this spell selection philosophy grew out of playing 3.5, where monsters never lived more than 10 rounds, the bodak that almost TPK'd us was retconned to being an illusion, and even high-mobility dragons refused to flee from combats they were clearly losing. In old-school play, it's very possible to engage a monster or group of monsters several times when one group or the other retreats, in which case a long-to-permanent-duration crippling spell against an enemy powerhouse could be very useful. Maybe there is a strong PC use-case for such spells in ACKS revolving around hit-and-run tactics. But my players are very... direct and true to their 3.5 roots in their tactics, and their spell selection preferences reflect this, with a strong preponderance of short-duration negative effects and direct damage instead. If I wanted to design a spell to fill the '3rd-level single-target offense' niche that Dismember is in, and I wanted to build something my players would love and use regularly, I would go for something like a single-target 1d8 damage per level blast, uncapped, no save, no attack roll, 90' range, which works out to 28 points under the spell construction rules, and hence a 3rd-level spell. Is this something which I would actually put in my game? Probably not, except maybe as treasure for killing a wizard who had it and was able to use it against the PCs first. And if some clever PC wizard decided to research it, I would likely have a rival steal his notes and use it against the party, because that there is a deadly weapon for the killing of dragons, kings, PCs, and other hard targets. There's a reason most damaging spells offer either an attack roll or a save for half; perfect reliability is both boring and powerful (especially when combined with uncapped scaling). Perhaps unfortunately, we seem to tend to gravitate towards it anyways...
(A more reasonable construction along similar lines might be a 3rd-level blast or death spell dealing 2d4 / caster level capped at 10d4, range 90', no save, no attack. 25ish damage is enough to put a really good dent in most targets, but it doesn't scale up indefinitely)