Not necessarily. If you can end things now, such as by casting Banishment on an extraplanar enemy, you save however many party resources would have been otherwise expended.
That's actually the same premise. If you can consistently do damage that will eventually equal or exceed the target's HP by 4 rounds but you can also cast Banishment with a 25% chance to hit, it may be more valuable to cast Banishment depending on reliability.
Now, if your chance to Banish was higher than 25%, then it may be safer to banish anyways, even if you have enough to survive 4+ rounds because you're more likely to save resources before it gets to that point.
Nearly every save or die spell is going to be con or wis save, with those con/wis save bonuses on top of the resistances it's safe to call bothering with a save or die spell a wasted spell slot. With 13 16 & 19AC, the fighter has three attacks each with a +10 to attack where he needs to roll a 3 6 or 9 to hit, calling them almost certain almost certain & 55% chance of 3x13 slashing for the fighter is fair. Disintegrate & chain lightning both use a spell slot & if not saved one will slightly* outdamage the fighter for one round while the other if not saved will somewhat fall behind. Except the level 11 caster only has one sixth level spell so can't cast both of those in one fight yet.
What you're referring to in the case of the Save or Die spells are what happens when the accuracy is different. When the accuracy of your random kill move is significantly lower than the accuracy of your consistent damage move, the damage move usually has priority.
Now, for the casting example, its true that AoE spells usually suck vs single enemies compared to single-target attack rolls from a typical melee combatant, but that's presumably on-purpose.
Well, first of all, we can actually determine the average number of creatures hit by an AoE using the DMG's formulae. For example, Fireball expects to hit (20/5)= 4 targets in range.
Now, I've actually been keeping an eye on how fireball compares to, say, a single-target character's attack. Actually fireball beats out a fighter's melee attack by alot, even accounting for GWM and extra attack...and I mean ALOT. This is assuming equal percent chance of success (I'll get to that).
Now, a fireball is more likely to kill any creature below 35HP, which is a pretty safe assumption against mooks. Above 35HP and the GWM fighter is more likely to kill.
But also, the average for Fireball is much, much higher than the average of GWM and it has less variance as well.
But what about unequal accuracies? Lets use the troll as an example:
+1 dex and 15 AC (84HP). Assuming level 6 and V.human , the fighter has a +8 to-hit. The wizard has a spell DC of 15, so the troll must roll a 14 or higher to save. In this particular instance, Fireball has the higher average damage, the lower variance, and has a higher kill chance for HP's below 17.
However, despite the higher average and other benefits of fireball, it turns out that above 17HP, GWM has a higher kill chance than fireball.