Seems to me, a/b/c do a lot of work.
Especially (b), ie the fact that spellcasting in D&D almost never requires a successful check.
Think about what, supposedly, the fiction of D&D spellcasting involves - precise hand gestures, speaking complex arcane syllables of such power and profundity that only a few of them can be impressed into a human brain at any one time (ie Vancian spell memorisation/preparation), pulling various material components out of pouches etc - not to mention the actual channelling and deployment of the arcane forces conjured up by the performance of these various acts. Then reflect on the fact that, per the game rules, this
is never mucked up!
Whereas a 1st level thief in AD&D has around a 1 in 7 chance to fall, quite probably to his/her death (d6 per 10' fallen on a d6 HD character)
every time the climb of a wall is attempted; and in 5e a fighter who wants to make a leap comparable to world jumping records (which are greater than 20 feet) has to depend upon a GM's decision as to the interpretation of the text (Basic PDF, p 59) that states that a STR/Athletics check "covers . . . try[ing] to jump an unusually long distance" - a bit of text that in a recent thread had a significant number of 5e GMs stating that the player would have to state some approach to making the jump beyond just
I give it my all, so as to warrant the making of a check rather than the GM just declaring automatic failure!
This is nothing to do with "each of the fiction and the mechanics informing the other". Nor is it to do with
it's magic - the
effects of a spell might be magical, but nothing in the game suggests that a 1st level MU has a magic ability never to drop stuff from his/her components pouch, or stumble or sneeze while speaking the words of power, etc. It's a
system issue, whereby a mechanical legacy inherited from Chainmail (of spells being fiat effects that a player can declare) is generalised across all fields of magical endeavour, while the system of ability and skill checks has been designed by grafting the mechanical example of attack rolls onto a half-baked task (as opposed to conflict) resolution system.
(Imagine if the sort of dexterity that 1st level MUs demonstrate in handling their spell components was extended to 1st level thieves trying to pick pockets!)
4e showed how the basic system conceits of D&D can be preserved and yet this particular problem overcome.
In "player-facing" systems, players who play martial characters KNOW FOR CERTAIN (before play ever begins) that (a) their conception of their martial character's thematic portfolio will coherently port from their mind to actual play and (b) they can reliably depend upon being able to change the gamestate and attendant fiction through that archetype manifestation as a result.
This is constantly underplayed by detractors of this approach, but it is definitely a thing for both long term players of martial characters who have been denied this in GM-mediated play (or at least rendered less secure) and in new players who look at their player counterparts who choose spellcasters and merely by din of doing so KNOW FOR CERTAIN (before play ever begins) that (a) and (b) will be realized because of the nature of D&D's supernatural-effect-by-fiat (I cast x spell vs some form of possible misadventure to spellcast because dice are rolled) inherent to spellcasting PCs.
Early on during the 5e/D&D Next design period, Mearls had an interesting blog about classic D&D fighters as "easy mode" and classic D&D MUs as "hard mode".
A different take on the same asymmetry is found in a c 1980/81 White Dwarf by Lewis Pulsipher, where he says that most serious D&D players prefer to play MUs (or something very much along those lines).
I've got not objection to a game being designed in this sort of fashion - different player options which set out different strategies with different difficulties of starting potential and different sorts of long-term destination - but one then wants the game to come out and say so! Gygax is a bit inconsistent in this respect. Some bits of AD&D suggest that all class choices are in some sense equivalent - it's just about
what function you want to play in the dungeoneering context (examples include the discussion at the start of the Character Class chapter of his PBH (p 18), and then the discussions fo class function that are found in both the PHB (p 106) and the DMG (p 86) discussions of gaining experience points and gaining levels). But other bits hint at the asymmetry - eg p 25 of the PHB says that MUs "are possibly the most fearsome of all character classes when high levels of ability are finally attained. Survival to that point can be a problem, however, as low-level magic-users are quite weak." (As far as my knowledge base goes, it's an open question whether this asymmetry was
intended, or simply
noted as a byproduct of developing the Chainmail mechanics so as to establish systems for playing both fighters and MUs as PCs.)
When we look at 5e, though, there is no hint of asymmetry between the classes. Nothing suggests that choosing to be a fighter is choosing to have fewer prospects of long-term flourishing within the context of the game. This is reinforced by the fact that low-level MUs are not distinctively weak as they were in AD&D.
It just seems irrational to think that you can seek to abolish this original asymmetry between the classes while keeping intact some of the key mechanical features - in particular, fiat spell effects - that underpin it.