Unfortunately this is not the case, because the mastery bonus is updated constantly and not snapshotted the instant you cast Meta. The same goes for intellect.

You can test this behavior by going to the target dummy, removing all your gear, casting Metamorphosis, re-equipping your gear (with a macro) and spamming Shadow Bolt until Metamorphosis fades. Do the same thing again, but without the gear swapping, and you will notice that the average damage done by SB will be very similar in both cases.

The following macro can be used to cast Meta and equip your gear (replace YourNormalSet with the exact name of a gear set in the built in Equipment Manager):
Do not swap trinkets when trying this, because equipping them triggers a cooldown on both their on-use and passive abilities.

Another, perhaps more definitive, test is to use a [Volcanic Potion], cast Metamorphosis, cancel the potion buff, and spam Shadow Bolt. Like in the previous case you will notice that the SB average damage is equal to an unbuffed Meta scenario.

When I tested this, a few weeks ago, the difference in Shadow Bolt damage between Meta casted w/o gear on and w/ gear was less than 1% on half a dozen attempts, so I concluded that mastery is not snapshotted.

I'll make a few more tests when I get home and post the results.

UPDATE:
I stand corrected, I make multiple tests on using Corruption, instead of SB, and mastery is indeed snapshotted when Meta is cast. So disregard my previous post.

Testing with trinkets, cloak and weapon not equipped so no procs can interfere with results.

Baseline (615 mastery rating):

Hitting meta and then applying corruption, corruption ticks for 1806.

With mastery elixir (840 mastery rating):

Hitting meta and then applying corruption, corruption ticks for 1838.
Cancelling the elixir so I'm back down to 615 mastery rating, and applying a new corruption before meta ends, the new corruption also ticks for 1838.

So meta still snapshots mastery when it's first activated, as previously tested.

Thanks Meegosh for providing some information of how much mastery gear that is obtainable.
Unfotunetly your equation to calculate the required mastery rating obtainted by a gearswap is incorrect. I'll share a better equation for determing what mastery rating you need to get with any given mastery rating in your "normal gear". A few minor assumptions will have to be made. Furthermore the assumptions are made in a way to favour not swaping the gear.

I'm assuming, just like Meegosh that you'll lose 2 seconds of metamorphosis due to the gearswap. Furthermore, I assume the 2 seconds to be "dead" in the sense that I'll do nothing during the 2 seconds. This is due to the fact that we'll likely suffer from gcd into the pull and it will obviously be less effective with 2 dead seconds rather than using metamorphosis "perfectly" hitting the boss at the very second we engage in combat.
I'm gonna go ahead and use the same formula Meatstorm used to express mastery rating required to gain 1 point of mastery. The value seems to be accurate enough even though it has the potential to diff a small amount.

PET = Dps of our pet.
DPS = Dps of you and your pet together
M1 = mastery rating in mastery gear
M2 = mastery rating in "normal" gear.
We will get the following expressions:

Total dmg with gearswaping for the duration of the first 36 seconds:
34 * (1.2 + 2(8 + M1/179.28) / 100 ) * (DPS - PET) + 34 * PET (1 + 2*(8 + M2 / 179.28) / 100 )

Comment: As you can see I'v taken into account that our pet will not gain any extra amount of dmg due to the gearswap. I'v also considered the 8 points of mastery we have regardless of mastery rating as well as the flat 20% from the actuall ability.

Total dmg without gearswaping for the duration of the first 36 seconds:
36 * (1.2 + 2( 8 + M2 / 179.28) / 100) * (DPS - PET) + 36 * ( 1 + 2 * (8 + M2 / 179.28) / 100) * PET

We want to determine when these expressions are equal in order to comupte how much M1 is required with given M2, DPS and PET. After some simplifying and solving for M1 we get:

M1 = M2 * (36 * 2 * (DPS - PET) + 4 * PET) / (34 * 2 * (DPS - PET)) + 1.36 * 17928 / 34 + 1.16 * PET * 17928 / (34 * (DPS - PET))

We can see in the equation that the less portion of dmg that comes from our pet the less mastery rating required to benefitt from swaping gears. If we'd assume just to get an idea how what kind of numbers we're dealing with here, that all dmg comes from ourself and none from the pet.
We'd get from replacing the approirate values: M1 = 717.12 + 18/17 * M2

Although this doesn't reflect a real fight very well. If we're using the latest simcraft for bis gear we can see that:
Simulationcraft Results
fellhunter = 7516 dps
dps = 28 709
mastery = M2 = 1053
Which inserted in above equation would give you a value of M1 = 2070.95 I.e. if you find mastery gear that gives you more than 2070.95 mastery rating granted that specific set of bis gear you'll gain dps from swaping gear.

The values I took from the simulation were from a fight far longer than 36 seconds. Hence the numbers are slightly inaccurate. Much likely both dps from warlock itself as well as dps from pet is quite abit higher and I admitt that you have to consider only the first 36 seconds average dps to get an exact value. Although I feel confident to conclude that given the fact we have such a great amount of margin to the calculated point of mastery where it's +- 0 that you in fact benefitt from swaping gear.

I think it might be possible to give a fairly good approximation of dps from the pet as a first degree function of overall dps. If it's true such an approximation could be made, the general formula for required mastery could be simplified alot.

Your equation is incorrect since you ignore the 8 point of mastery you have per default as well as the 20% dmg gained from the ability metamorphosis. Furthermore, using your assumption that you don't suffer from gcd into the pull (i.e. using metamorphosis at t<0) you need to consider 2 more seconds of a damage multiplier 1 (nothing).

Just taking the first two mentioned flaws in consideration which is the reality you aiming to adress we'd get:
36 * (1.32 + 2 * M2/17928) DPS = 34 * (1.32 + 2 * M1/17928) DPS =>
2 * 1.32 + 2 * 36 * M2 / 17928 = 2 * 34 * M1 / 17928 <=>
M1 = 17928 * 2.64 /68 + M2 * 36/34 <=>
M1 = 696 + 18 * M2 / 17.
For M2 = 2465.2 we'd get M1 = 3306
M1 would in fact be slightly lower though given we should as mentioned consider 2 more seconds of dealing dmg outside of meta.


What I'm trying to say with 2 dead seconds is that in the intervall 0 < t < 2 I'm not doing anything. The reason I did choose to make such an assumption is partly because granted we'd use the macro to use meta and change gear at exactly t = 0 we would in fact suffer from a gcd into t>0. The alternative would have been to make the perfect scenario assumption that we'd start to dps at t = 0 with a loss of approximatly 2 seconds of metamorphosis. Since the idea was to prove it's actually benefitial to use gearswap I went for the option which would cut back the most on the dmg.
In your equation you're actually doing the very same thing. Considering 2 seconds of "dead" time, although given your response it wasn't perhaps your intentionally done.

Furthermore I thought it was a fair assumption we wouldn't engage our pet at t = 0 even though it's possible. I believe I made the assumption up because it would better reflect my very own likelyhood of human error when I'm forced into timing metamorphosis. I'll have to agree from a purely theorycrafting perspective such an assumption is stupid. Either way with that assumption being made one would have to consider pet dps and personal dps. But granted we're engaging the pet just as quick as if we wouldn't have been gearswaping, you don't have to take it into consideration at all and the equation stated above is the approiate one to use.

It's sort of funny to see how much of an impact a relativitly small human error has though.

Well, I did take the 8 mastery points and the 20% meta into account, you just seem to have not understood it : the 36 in the first equation represents the 20% meta basic damage (20) plus the 8 basic mastery points (+ 8*2). I believe that what fooled you is that I maybe was a bit hasty, and had allready simplified both equations by 1/100. Consider also that I wasn't interested in comparing the damage of the 2 executions, I only considered what you gain (34 seconds of boosted meta) and what you lose (2 seconds of meta, not boosted) which is not what you had chosen to consider (both approaches are valid though).

Considering the 2 seconds issue, I have no problem considering the worst case scenario in all that we can't really control, but your solution here (starting dps at t=2 instead of t=0) is not really necessary (I don't see how difficult it is to meta before the pull and start dps at t=0) and it really has a great influence on the results (not "slightly" as you said, since M1 goes way down to 2800 in your exemple if you consider a t=0 dps start).

My apologieses. I was under the impression from first looking at your equation that you were considering diffrent conditions than stated. Looking at it again with your previous post in mind I can see that it's true starting at t=0

You're right: the standard simcraft profiles (for the normal dps measurement, or for the trinket comparisons) do not try to take trinket procs into account in their priority lists. It's quite easy to modify the priority list yourself to test this out, but the proper strategy for using it should probably also factor in the ICD on the mirror.

As simcraft has an average proc interval of 114 seconds, and the proc itself lasts 20 seconds, this gives us a reasonable chance of lining up enough Impending Doom procs to use a meta each time it procs, in which case it's likely to be very valuable. However, if you miss a mirror proc then waiting until the next one is very likely to end up losing you a Meta, while popping Meta early is likely to make you also miss the next mirror proc. This strategy is likely to perform very well when you're lucky, and very badly when you're unlucky! It would be possible to decide whether or not to wait on a mirror proc based on your knowledge of when it last procced, but it would be difficult to determine where the appropriate cutoff actually is.

A reason you may be beating Simcraft at the moment is that you're significantly better geared than the 359 profiles currently in use? Also, don't forget that if you're raid force has a high dps and the fights are short then your DPS will be better than that simmed for a longer fight. Ignore this is you're running the sims yourself with your own fight parameters though, in which case you're strategy (and indeed your play!) must be extremely good.

As far as I know, Meta does not affect our pets or guardians, as they have never scaled from direct percentage damage increase, only from stats (if they scale at all, and guardians such as the Doomguard snapshot when spawned for their scaling). However, something strange is going on...

Your Doomguard on Chim is doing average 35k hits and 70k crits, well over triple the damage caused by the Simcraft one, also, he only managed to cast 20 times compared to the sim's 32. I'd say that the simcraft model for the Doomguard is not matching up with the play experience for either the cast time of Doom Bolt or the damage it causes. Both this and the shorter fight length will greatly increase the proportion of your damage that comes "for free" from the Doomguard versus the standard length Simcraft fight with the possibly incorrect Doomguard.

Well, first, the easy thing : the Doomguard isn't affected by +dmg buffs, even your meta. However, the mastery boost is huge for him, since it also boosts other intel/sp buffs.

Now for the hard part. What we want to know is the probability of our meta being up before, or during, the trinket's proc. I stated before that I though this trinket was a waste, mainly because of RNG. Well, let's look at it a bit more :

1)We need to calculate the average time between the proc of the trinket and the fading of it's next buff = T

2)We need to calculate the probability that our meta is available at T = p


1) T = t1 + t2 + t3, with t1 = the trinket's ICD = 100 seconds, t2 = the duration of the trinket's buff = 20 seconds, t3 = the time it takes for the trinket to proc. For calculating t3, we will consider a "normal" situation and a 1993+ haste rating : immolation, corruption, bane of doom ticking, shadowflame on cd, shadow bolt casting :

immolation : 0.42 ticks/second
corruption : 0.42 ticks /second
bane of doom : 0.07 ticks/second
shadowflame : 0.3 ticks/second
shadow bolt : 0.45 ticks/second

Hence we have 1.66 ticks or casts per second, each with a 10% proc chance. Thus, in average, t3 = 4.2 seconds and T = 124.2 seconds.

2)We need to reduce the cooldown of meta by 55.8 seconds, so, we need 4 procs in these 124.2 seconds. Let's calculate the probability to get 4 or more procs in these 124.2 seconds. For that, we need X, the number of shadow bolts/incinerates/Hands of Gul'Dan cast in that time. Let's call Y the number of procs :

p(Y = 0) = (X 0) * 0.85^X = 0.85^X
p(Y = 1) = (X-1 1) * (0.85^(X - 1)) * 0.15 = (X - 1) * (0.85^(X - 1)) * 0.15
p(Y = 2) = (X-2 2) * (0.85^(X - 2)) * (0.15^2) = (X - 2) * (X - 3) / 2 * (0.85^(X - 2)) * (0.15^2)
p(Y = 3) = (X-3 3) * (0.85^(X - 3)) * (0.15^3) = (X - 3) * (X - 4) * (X - 5) / 6 * (0.85^(X - 3)) * (0.15^3)

Thus, p(Y >= 4 ) = 1 - (p(Y = 0) + p(Y = 1) + p(Y = 2) + p(Y = 3))

With X = 30 (medium movement), p(Y >= 4 ) = 74%
With X = 40 (very light movement), p(Y >= 4 ) = 89%

Let's now cover the best case scenario, where your meta is up before the end of the ICD (in that case the next metas will be way easier to line up with the trinket).We have then T = 100, and we need to get 6 procs.

With X = 24 (medium movement), p(Y >= 6 ) = 45%
With X = 32 (very light movement), p(Y >= 6 ) = 59%

Hence we can see that on a relatively still encounter, we have 89% chance to get the second proc and 59% chance for the situation to not deteriorate. The chance for us to get all of the procs is a little more tricky, here is one estimation though : on a very light movement encounter where you use meta 3 times, you have at least (0.59*0.89) = 52% chance to use all metas on proc. For 4 metas, 31%
You can change these numbers with your personal Xs according to the encounter.

Take these numbers as you wish, I personnaly believe though that on medium to heavy encounters, you really shouldn't use Theralion's mirror since you'll probably miss the third proc and maybe even the second, it is however indeed interesting on light to no movement encounters.

To my mind, the main drawback of this trinket, aside from its RNG, is that to get the best out of it, you have to meta on the proc, hence you may not be able to meta when you want, more precisely on encounter phases where you want to have the most dps (Chimaeron's second phase, Cho'Gall's second phase, Ascendant Council's third phase ...) or miss the proc to keep your meta, rendering the trinket quite useless.

It seems that SimCraft indeed is calculating Doomguard DPS wrong. It has a casttime of 3 seconds regardless of haste, which is (1.08 * 60 sec) / 3 = 21 casts which usually is the case for demo. Obviously fewer casts as destro and affliction. The demonology BIS profile calculate it at a ~2 sec cast time, which is not possible.

Unfortunately I've been unable to import my character into SimCraft (for some reason it doesnt accept me reforging crit_hit on Sinestra head).

Either way it's a significant DPS boost popping Doomguard at the start with all your procs and volcanic pot as this is generally the only part in the fight where you have all your procs lined up at the same time. It accounted for 14% of my DPS on Chimaeron, albeit with the mirror proc.

Obviously a small sample, but the other lock in our raid popped Doomguard at approx the same time with the same procs up (bell instead of mirror is the only difference). My Doomguard did 4k eDPS compared to his 2.6k eDPS, which is a 1.4k eDPS difference on a 275 sec fight. His numbers match mine when I use bell instead of mirror. That is, in my opinion, a pretty significant DPS boost for a trinket proc. Probably needs a lot more testing though.

Lastly, I have to agree with what the poster above me stated, and like I said in my previous post; The trinket relies heavily on RNG. It seems to work perfectly fine on Patchwerk fights, which I guess mainly will be Baleroc in Firelands (where you might want to spec Destruction for Nether Ward).

I must say I disagree on the trinket being a waste though. It's still ranked pretty high despite SimCraft not popping Meta during the proc.

The OP lists a belt, [Cinch of the Flaming Ember], as an alternate to the [Reckless Ember Topaz] in a yellow gem slot. Which would be cool but would require some pretty intricate origami.