Looking at notorious periods of imbalance from when WoW still thought of itself largely as an RPG (where Rez was a utility of convenience, alike to summoning stones, mage portals, and even being merely able to spec into a tank, and the "game" and its "content" was not yet limited purely to "endgame") to make a point for the necessity of homogenization is about as disingenuous as discussing the sustainability of nuclear reactors on the sole basis and sole knowledge of a nuclear bomb. Early WoW and XIV are, yes, both MMOs. Their applicable comparisons end close behind that, however, unless you wish to discuss the benefits of world-oriented approach vs. endgame-oriented approach to design, where it can at least have tangential relevance, or the appeal of particular settings, archetypes, or tropes (as per any MMO or work of interactive play). We expect better polish from this game so deep into the genre than we did from, say, the Burning Crusade.
You can have job identity without being denied basic tools. You can have job identity while still having the ability to control threat/enmity. In fact, you can even do so in unique fashion, so long the basic capacities are met. You can absolutely make up each and every necessary capacity of a tank without needing identical gameplay or mechanics.
It's easy to cherry-pick non-examples from one setting and toss them verbatim at another as if they implied something; the gaming world is filled with failures and incomplete or flawed works to choose from at your leisure. But if you can't show exhaustively how every homogenized mechanic among tanks absolutely demands that exact approach to meeting the capacity it suits, then you have no actual concept by which to find homogenization necessary.
Tanks can function diversely and feel individually unique and still be balanced. On the other hand, merely ensuring that every tank has the same base tools and then expanding there-past on the assumption that their shared core will give them balanced value in content is foolish; it guarantees nothing except that they will start with the same capacities and that their uniqueness will always be fettered to that shared block of "Tank", rather than "Dark Knight", "Warrior", "Paladin", and "Gunbreaker".
Food for thought:
Paladin could at one time Cover in order to essentially double-Hallowed Ground crucial mechanics--such as during pre-fix Firefall, pre-buff stacks, or the orbs on Omega Ex, allowing a minor sort of cheese strat for struggling parties. This wasn't needed but once, but it was damn powerful.
However, when this was fixed, rather than seeing Paladins fall off with their greatly diminished value (though, they were already basically 50/50 on average), Warriors started being benched. You see, once the Paladin could no longer cover for both tanks in order to solidify the swap or deal with both the TB and orbs simultaneously while healers were both frantically keeping the rest of the party alive, they simply started taking two Paladins. The noted imbalance fix had backfired because it actually restructured the fight design and tightened the apparent need for a second mobile immunity skill.
As you've pointed out, the fit of a job or class into a given fight has often less to do with the job itself than the mechanics of the fight. But there are periods where a particular (excessive) power acts not in the advantage of the job itself but for another because there is only so much to cheese or so many checks to fill. Make a ~6-minute fight have just one thing to immunity, but through which only Hallowed Ground can last, and you force Paladin. Give the same fight a frantic triple-crit flurry that would otherwise require no immunity, but would require immense healer attention, and you greatly entice use of Dark Knight for its pre-immunity stage (more so if the skill were more functionally balanced). Give an add phase damage enough to entice an Immunity in place of a CD during a less hectic, but significant, tank-flurry preceding it (repeated twice or thrice over the whole fight) and you entice use of Warrior.
No matter what you to the jobs themselves, those jobs variances will have to be accounted for. But, you'd be amiss to guess that just because one shrinks the differences between kits, the differences in community performance expectations would shrink as well. It's often the opposite. Give nothing worth experimenting with, and community trends tighten and narrow even faster. Prove through your fight designs that you will give something worth experimenting for, and community expectations remain open far longer. That confidence experimentation can be rewarded--precisely because fight designs show that they were made with jump times and tankbuster timings and mobility periods with the job toolkits in mind--often does more to let players enjoy the jobs of the game (and not just they themselves play) than any ease of optimization.
Sidenote: I had no significant troubles co-tanking with Warriors who had no Provoke. Admittedly, though, it was much easier if they had XIVapp for its enmity tracker, as that function hadn't been added to the official UI yet when I ran into most of these no-Voke Warrior mains. It was easier, of course, if I was a Warrior as well or I could get significant life-saving value of Stoneskin, but alas, I'd never failed a meaningful swap for it outside of one Ifrit Extreme run. Likewise, I had no difficulties tanking on Paladin without a direct taunt (HotR); if anything my pull-three-from-target-ally functionality was better for speedpulling multiple weak mobs from patrols at once via a Feign-Death-pulling Hunter, as I would only keep the three and all others would be returned to where they were pulled from.