Which edition best reflects the archetype, as opposed to which one does the mechanical class best? Well, in general it'd be the very early editions - OD&D, B/X, 1e - mostly because since then archetype reflection has been in many cases somewhat sacrificed on the twin altars of balance and playability. Further, some classes have yet to get it right at all despite repeated attempts. My list, including some classes not in the original poll:
Cleric (Normal) - 1e maybe, as the best of a rather poor lot. A properly-designed archetypal Cleric should (unfortunately) be every bit as annoying and hard to play as a Paladin, only from different points of view: righteousness has 9 alignments.
Cleric (War) - we homebrewed this many years ago but official D&D has never given us a true Battle Cleric archetype beyond some 2e specialty kits. Still waiting, as there's a fantasy archetype there to be filled.
Druid or
Nature Cleric - hasn't been done, as the archetype would likely not be an adventurer as required by the game.
Fighter - very dependent on what one sees the archetype as; there are many. 3e could build a good swashbuckler archetype, 0-1-2-3-5e a good simple front-line soldier, 4e had the front-line stalwart defender, and so on. Pretty much any Fighter one can build in any edition is going to more or less replicate some archetype or other - hard to go wrong.
Paladin - 1e set the standard, but - like it or not - almost any edition gets close enough.
Cavalier or
Knight - a very important archetype, distinct from a foot-soldier Fighter. 1e's Cavalier - meh. 3e's Knight had potential. Difficult to do with an ordinary Fighter.
Ranger - Aragorn's the archetype, so 1e, except for some silly restrictions e.g. alignment. Anything post-Drizz't is garbage.
Barbarian - shouldn't even be a class IMO but if it must be, 3e is probably the closest I've seen to an archetypal Barbarian
Wizard (Magic-User) - 1e or early 2e. After that they become the poster child for sacrificing archetype for game play.
[Wizard subclasses or specialists] - the only ones worth the effort are Necromancer (could easily be its own class) and Illusionist. Neither archetype has really been done well yet.
Sorcerer - 3e.
Warlock - not even sure what archetype this is going for, so...whatever.
Psionicist - hasn't been done well yet and maybe never will be, but if someone could design a class that can work like Kurtz's Deryni I'd be very interested...
Rogue (Thief) - 1e for the archetype of a sneaky but daring (yet somewhat fragile) Thief e.g. Locke Lamora. 5e is getting there. The archetype of the dashing swashbuckling "rogue" e.g. Jack Sparrow is otherwise best served by building a light-armour high-dex Fighter.
Assassin - another one where the archetype has always to give way to balance and game play. An archetypal Assassin works (mostly) alone and either shoots from range or sets up something deadly (poison, trap, bomb) and leaves; none of which works well in a party/group-based game. So the game tried to make them more like Ninjas...which would be fine except for the existence of the Monk class, which does it better...or dangerous sneaks; but we have Rogues/Thieves for that.
Bard - has never been done well and at this point I'm prepared to say never will be. Would likely work better as a non-adventuring class similar to Sage; sure the archetype is a wanderer but not one who gets into dangerous situations - instead they tell tales about other people getting into dangerous situations - and an information source.
Monk - though we can argue forever as to how well they fit in to the quasi-Eurpoean fantasy milieu, there's no question the sort-of-Ninja archetype has been reasonably well-served by the class in all the editions.
Lan-"a disciple of the bad-ass might-is-right Fighter archetype"-efan
