If this is correct then with:

Parry: 17%
Dodge: 12%
Block: 26%

You end up with 17+12+26 = 55% of the table/pie already filled. Presumably all you need is an extra 45% block chance (or even less because I haven't taken into account M) to knock the rest like glance, crush and crit completely off the table? That is far lower than the +75% block chance you get from activating the warrior shield block skill. Am I understanding that correctly?

If that is correct then in order to be immune to crits and crushing blows with shield block up you just have to scrape together enough parry, dodge and block in total to hit 25% (less in fact when you consider that M would account for a certain percentage too). You could manage that quite easily with fairly poor tanking gear. It also doesn't tally with the idea that 25% base block makes you immune to crits with shield block up.

Thats right.

I might have missed something, but I'm trying to understand the attack roll table

example; A rogue with +13 to hit and +30 to crit means your table will look like this on a lvl 63 mob;
11% chance to miss (24% base miss rate with 2 weapons) + 15% Dodge, parry, block + 40% glancing blows +30% crit + 4% regular attack damage

Is this how it works?

This has always irked me and there's no reason why it shouldn't be changed. Seeing white resists in PvE with huge +hit, especially against mobs that aren't even level 60, is just dumb, while in PvP it can mean the difference between winning and getting your skull crushed.

Wait, then, I just realized--of what significance are Glancing Blows against level 61?

Thats correct, increasing your chance to block has no bearing on how often you are crit/crushed unless you are wearing very bad gear.

Mocking blow tooltip says it generates a moderate amount of threat. Not at home now so I can't double-check it but i'm 95% sure.

Most of this is nitpicking, but...
I think it's more likely that most buffs generate threat, but such a small amount that it doesn't affect normal combat, only the pull. I will claim without proof that bloodrage / berserker rage / battle shout generate threat, but shield block or a HoT (before it has ticked, just the cast) wouldn't pull a body-pulled mob - they truly generate 0 threat.

Actually, i did test this once. Get X to body pull, then Warrior to taunt and do nothing. At the end X will have aggro, but Warrior is clearly on the threat list. Now if Warrior casts shield block, he doesn't pull aggro.

As regards the threat stacking thing, my research has found that some bonuses do stack additively. These are, all together,
Berserker Stance / Battle Stance / Rogue Stance: -20%
Tranquil Air -20%, Blessing of Salvation - 30%
Defiance / Feral Instinct +15%
Arcanist 8/8 -15%

Most of the other bonuses are multiplicative, including nemesis 8/8, netherwind 3/8 (i.e. it acts before the other mods), arcane subtlety, random talent X.

e.g. a Rogue with BoS will generate 50% threat per damage, allowing him to do 60% more damage than without it!

e.g. Mage + BoS + AS + 8/8 Arcanist = (100 - 15% - 30%) * 60% = 33% threat. Well, not for long, thanks to 1.11.

Check this. I found 1 point of healing was 0.25 threat for Paladins.

My experience is that if gives you the threat of the player who has aggro, not the highest. (Otherwise the MT could use it to increase his threat if people were above him).

Yeah, that makes more sense (also, I think Bloodrage and Battleshout have recorded threat levels). It seems pretty likely that Fire Ward or PW:Fort or all the other silly things people can do to pull aggro from Firemaw as he's flying towards the raid all have measurable threat levels.

It's good to know that there's no specific rule in place making mobs go after the most recent 0-threat ability caster in a situation like that. Maybe when I'm less busy I'll make a list of things which are safe to cast during a pull.

Actually, as of right now, we are getting flat aggro redux in 1.11. :dance:

Huh, well, that's odd. You're definitely sure about BoS stacking with, say, Arc.8/8 (the data was good enough to resolve 0.55 threat from 0.595)?

Okay. The 65% number was pulled off of this board, quite a while ago. Still, 50% threat? With BoS, that makes the 1.9 Pally threat "nerf" pretty meaningless.

Right, that was just lazy wording.

It's still odd how it works on rooted mobs (Rag considers the high-aggro Hunter his current target for the purposes of Taunt). Does the 10% rule even come into play in a situation like that (where the mob isn't attacking the person who he'd like to be)? Might be good to know if they ever introduce another rooted mob who melees.

Could I get BoP at the beginning of a fight, blow all my timers, and then have the tank Taunt it "off" me (even though it's actually been on him the whole time) before BoP wears off, thus getting a heap of free threat? Seems exactly like the Rag situation.

Here's my data for Tranquil Air + Arcanist 8/8:

Shaman does 187 damage to mob, then stops (no TA down). Aggro will change after 187 * 1.1 = 205.7 threat.

Mage has TA up. Mages does 315 melee damage (no aggro), then a hit for 3 damage (gets aggro).
Therefore his modifier range is from (205.7 / 315) to (205.7 / 318) = 64.7% - 65.3%.

This is in range of 65%, and not 85% * 80% = 68%, which supports additive.

Interestingly, thottbot seems to know which ones are additive and which not. For the additive abilities there is the effect "Apply Aura: Mod Threat", whereas the multiplicative abilities are "Apply Aura: Add % Modifier (Threat)", i suppose these are kept in the MPQs somewhere.