First, excellent analysis. Second I'm going to disagree with this part in specific.
Resolving 2 or 3 attacks/round with the same target number and same decisions being made doesn't IME take 2 to 3 times longer than one attack. It might for a computer - but for a player you only need to check the target number once (for both rolls) and you can easily grab two different coloured sets of dice. And it's one thought process. It takes longer than one attack, but not much longer.
It doesn't have to take 2-3 x longer, it just has to take
longer.
A round where you roll to attack (possibly with advantage), then work out how many hits, then roll damage for however many hits land, is slower than one where you only do it once. Even if you do everything
right and efficiently (colored dice, etc).
A OD&D fighter attack might look like this:
d20 - 3, hits AC 9, 1d8+2 - 5 damage.
A 5e paladin attack with nothing really special going on might look like this:
d20+13 (17) hits AC 30, 1d12(4)+1d8(6)+7 = 17 (6 radiant)
d20+13 (5) hits AC 18, 1d12(10)+1d8(1)+7 = 18 (1 radiant)
ok first hit, second missed, so 17 total (6 radiant).
Even roll 20 doesn't make the second as fast as the first.
The thing is, the change in the state of the fiction between the two mechanics
is the same. You went in there and wounded a big monster, and the big monster didn't die.
Possibly the 5 damage was the same percentage of the monster's total HP as the 17 was as well.
And going from 1 to 2 attacks? Might take 25% longer. 2 instead of 1 digit modifiers, an extra damage die, multiple damage types? Another 10% longer. Going from 2 to 3 attacks, another bit of slowdown.
The fiction-per-second goes down, which slows combat down. Optimizing for fiction-per-second is valuable. And in 5e as character power accumulates, so do mechanics, and those mechanics slow down the fiction.
3.X and 4e made different versions of the same mistake here. The 3.X iterative attacks mean that you've got to do linked but different calculations for each attack because there is at least one extra modifier on each attack (plus a different one on TWF). The 4e standard actions are fine for speed (as long as you don't go over about three or four dice; you're right that this is a problem at 7d12). The problem comes when you're using minor action or interrupt attacks. These are added processes and generally also have less familiarity (which speeds any process up, so although the mechanical actions between a two weapon fighting off-hand attack and a minor action attack encounter power may bethe same you'll always be doing the TWF).
The interrupt attacks are a huge load -- because they interrupt other turns, breaking the flow of the game and making it take longer to get around a round.
4e made multiple smaller taps far more effective than one large tap because you could stack static per-tap damage. 4e also made HP grow fast enough that the "big taps" at high levels where not big enough; a 7d12 attack by the time you got it didn't
do anything impressive.
3 1[W] attacks did more damage than a 7[W] attack did, but took longer to resolve. And the 7[W] attack had less fiction impact than a 3[W] attack at level 1, but also took longer to resolve.
The "optimal" 4e weapon user had enough encounter and at-will powers to make 3-4 attacks on their turn then 1 interrupt, at least for the first 2-3 rounds of the game. This far outdamaged any "big hit" build, mainly because 4e didn't have many optimization opportunities to boost a big hit, and had plenty for small repeated hits.
Crit-fishing was the closest to optimizable "big hit" and it also relied on rolling a lot of dice. One avenger I played had an encounter power that did 4 attacks, each rolling twice, each hitting for 2d6+big static, a 19-20 crit range and a fistful of crit dice.
An attack round could easily involve rolling upwards of 30 dice, and as fast as I could roll and scan and the like, this took time.
The fiction impact of that? Decent, but it was mostly a single number (damage) reduced from the enemy's HP. (the number was often "all of it").
All of those additions, conditions, rolls, checks -- they where
mechanics that didn't directly impact the fiction.
The other thing is the way the hit points escalate with level. 4e hit points go up linearly. 5e's I would say go up faster - but whatever they do the CR charts in both the DMG and XgtE are utterly broken.
4e monster HP where (L+3)*8 or so.
The DMG charts aren't as far off as you think. They are tools for calculating CR from a monster, not per-level advice.
Look at bruisers -- monsters that do little other than soak up damage and deal it. They'll match the CR charts far closer. Even then, because CR is the product of offence and defence, a monster will often be way higher in defence or offence.
Then you have to factor in the myriad special defences CR11+ creatures usually have.
I hear people talking about how CR charts are broken, but when I take a non-cherry-picked monster without lots of weird abilities from a MM and paper-napkin crunch it through the CR calculating algorithm, I don't get a value that far off from the MM value.
What more, if I take the CR algorithm and I reverse engineer what I suspect they are doing "behind the scenes", I get something with reasonable amount of mathematical elegance; XP is the primary unit it appears. Take HP times DPR, throw in a modification factor for AC/ATK, then multiply it by a constant (there may also be an exponent in there, I forget off the top of my head), and you get XP. And then CR<->XP is pretty close to a constant and exponent relationship.
The exponents involved are close to the exponents that mimic the "number of foes" scaling on XP, which calls back to the military tactics rule of how armies scale (which is between linearly and quadratically; the first is how armies that fight 1:1 scale, the second is how armies where unlimited "ganging up" can occur scale).
The 5e designers did the math, they just didn't make it as obvoius as the 4e designers did.
But there's one final major missing difference. And that's level range. oD&D was basically a level 1-10 game with a soft cap at L10 where player hit points slowed down. WotC D&D has levels up to 20 in play - and both 3.X and 4e have an epic tier with ridiculous modifiers.
As mentioned, OD&D fighter-type damage output relied on more accuracy climbing instead of damage per hit or swings per turn climbing.
There is a limit on how much accuracy you can add (eventually you always hit). So more meat can't be compensated for with more accuracy one you are almost always hitting.