Thanks Mode, I'll take that as a word. People around here seem rather attentive anyways.
So what I did is basically nothing revolutionary, I simply put some figures together to see how exactly parry influences melee damage on the tank and to what extent expertise has influence on this. Also, some extreme forms of parryhaste may always occur and, being both highly unlikely and not linear, the chance to suffer such a scenario is lowered in another way than the overall dps reduce by stacking expertise.
Factors I took into account were:
EP - Enemy Chance to Parry
SE - Enemy Attack Speed
SP - Player Attack Speed
G - Use of GCD for attack that may be parried (this makes up for the difference between warriors, paladins and druids)
I tried taking player's own parryhaste into account and the result was so minimal, it's absolutely not worth the loss of clear relations.
The figures I took from placing these factors into relation:
AR - Player Attacks per Enemy Weapon Swing
AR = SE/SP + G*(SE/1,5) *fixed
PC - Effective Enemy Parry Chance
PC = EP*AR
PD - Effective Chance on Enemy Double Parry
PD = EP*EP*(AR-1)
PH - Effective Enemy Parryhaste
PH = PC*0,24 + PD*0,08
PD is less effective for PH than PC due to a much larger chance of a second parry during the same swing landing in the part which grants less to no haste bonus. After all, it might have a little bigger impact, but it's also not worth the possible loss of clear relations.
Now I present the results for different sets of input:
EP: 10% EP: 10% EP: 5%
SP: 1,5 SP: 1,5 SP: 1,5
SE: 2,5 SE: 2,5 SE: 2,5
G: 1 G: 0 G: 1
AR: 3,33 AR: 1,67 AR: 3,33
PC: 33,33% PC: 16,67% PC: 16,67%
PD: 2,33% PD: 0,67% PD: 0,58%
PH: 7,81% PH: 3,95% PH: 3,95%
As you can see, not only is parryhaste a considerable increase in Boss dps on the tank, both boss dps and chance on double parries by the boss can be decreased significantly by either stop using your GCD for attacks that may be parried or by reducing the enemies parry rate by 5%.
This proves true for other enemy attack speeds as well:
EP: 10% EP: 10% EP: 5%
SP: 1,5 SP: 1,5 SP: 1,5
SE: 1,5 SE: 1,5 SE: 1,5
G: 1 G: 0 G: 1
AR: 2 AR: 1 AR: 2
PC: 20% PC: 10% PC: 10%
PD: 1% PD: 0% PD: 0,25%
PH: 4,72% PH: 2,4% PH: 2,38%
EP: 10% EP: 10% EP: 5%
SP: 1,5 SP: 1,5 SP: 1,5
SE: 3,5 SE: 3,5 SE: 3,5
G: 1 G: 0 G: 1
AR: 4,67 AR: 2,33 AR: 4,67
PC: 46,67% PC: 23,33% PC: 23,33%
PD: 3,67% PD: 1,33% PD: 0,92%
PH: 10,91% PH: 5,49% PH: 5,52%
Looking at this data, we can see that the chance to trigger a double parry on a boss can be reduced to a chance below 1% in most cases by either simply stopping attacks on the GCD or reducing the parry chance of the boss by 4-5%.
I have made calculations for a reduction to 2% parry chance as well and even though the overall incoming dps reduction seems somewhat linear, reduction on the chance of double parries is not that strong. Very slow boss attack speeds being an exception, since many player attacks land on the boss during one swing and therefore even a minimal parry chance seems to have still some kind of impact:
EP: 2% EP: 2% EP: 2%
SP: 1,5 SP: 1,5 SP: 1,5
SE: 1,5 SE: 2,5 SE: 3,5
G: 1 G: 1 G: 1
AR: 2 AR: 3,33 AR: 4,67
PC: 4% PC: 6,67% PC: 9,33%
PD: 0,04% PD: 0,09% PD: 0,15%
PH: 0,96% PH: 1,59% PH: 2,23%
So the obvious benefit from stacking expertise (except of course increased threat generation) is also a considerably smoothed out boss dps. Reducing burst damage spikes from 2,3% down to 0,6% over the course of the fight seems to me like a big advantage. Even more so when fighting bosses with slow attack speed.
My personal opinion resulting from these numbers is that stacking expertise up to the dodge-cap is strongly recommended for every tanking warrior.
Concerning druids and paladins, try altering the G variable to a more realistic value.