EH and MEH do not need to be homogenous, but they do need to be within a reasonably small delta. It's okay if one class has a slightly better overall EH/MEH and another class has slightly worse overall but better for a specific sort of damage or better when damage can be predicted for cooldown use.
But that swing can't be very large, or stuff gets nasty, as has been discussed.
I still think that +/- 5% is a good rule of thumb. That would mean that a class with good general EH should have about 5% better than a class with good targeted EH or cooldown EH. Targeted EH could boost the EH of a tank by about 10% (i.e. "I'm good at taking magic damage" or "I'm good at taking melee damage.") from the poor EH point. Cooldown EH should probably *average* to a 10% swing (i.e. up for 1/4 of BURSTS means a 40% boost).
Something along those lines, anyway. Then you can have the following scenarios:
Mixed damage, unpredictable bursts: an average EH class has a 5% advantage over the worst.
Physical damage, unpredictable bursts: a physical EH class has a 10% advantage over the worst, and a 5% advantage over the average.
Magical damage, unpredictable bursts: a magical EH class has a 10% advantage over the worst, and a 5% advantage over the average.
Mixed damage, predictable bursts: a general cooldown class would have a 40% advantage on one quarter of bursts over the worst.
Magic damage, melee damage, etc.: Same dealio.
I think one thing with the setup we have now is that cooldown uptime and mitigation amounts appear to be based on overall uptime, and not percentage of bursts they can catch. It's also a bit tricky because there should be some skill involved in choosing when to use cooldowns--which probably means a larger EH boost than would normally be indicated, under the assumption that CDs won't be used every time they're available and the burst occurs.
In any case, if a given class has more than a 10% EH advantage overall on a given burst type, that class is likely to be considered mandatory for fights of that nature. 10% is already large enough that a major progression guild would always use a tank of that type, and less "leet" guilds would prefer a tank of that type if they have one. But it's not so large that a guild would say "We just can't do this because we don't have an X on our tanking team." More than 10%, and you're really starting to get into that area of people not even wanting to attempt the fight with the "wrong" kind of tank.
Note: cooldowns can also be tailored to different bursts damage types, of course. So you could imagine a class that's generally strong against melee and weak against magic, but has mid-grade general damage cooldown abilities. In fact, that would probably make good sense: have classes vary in magic/melee/mixed EH between "-5%", "average", and "+5%", and magic/melee/mixed cooldowns between "none", "moderate", and "strong".
For extra credit, make some of this selection based on talents that are difficult to take all at the same time as other desirable talents. So, for example, an anti-magic specced character might have weak physical cooldowns and strong magic cooldowns, but an anti-melee character might have moderate general cooldowns and better overall physical EH.
Then you can tweak the actual choices of what each class should be strong and weak against so that nobody is a "do-everything" class.
ANYway, I've gotten pretty far afield with speculation about how to design a game with a balanced but rich selection of tanking options. Still, I think that the very core of my thesis holds strongly: no tank should have more than a 10% EH advantage before cooldowns over any other tank for either magic or melee damage. Cooldowns and advantages against specific sorts of fights can add flavor after that, but the core EH can't vary more than that much without one class being "best".
Learn to science and stop theorycrapping in its tracks.