Actually, it looks like the optimum is no Double Strike/Thrash at all,

*if* the goal is to achieve TkT as frequently as possible. I had a bit of a play with function optimization:

http://imgur.com/VDojU
Code:

NMinimize[{((x + y + 3) / 2) + 3, x >= 6 - 1.5(0.4931 * z) && y >= 7.5 && z > 0 && (23 / z) + (39 / x) + (30 / y) + (30 / (((x + y + 3) / 2) + 3)) + 0.5495 + 1.1111 < 13.4162}, {x, y, z}]

In this context, x is the average frequency of Project/Shock, while y is the average frequency of Slow Time/Wither and z is the average frequency of Double Strike/Thrash. Kinetic Ward/Dark Ward and Force Breach/Discharge are considered. The assumption is that DS only reduces the timing of Project/Shock by at most one GCD. It's a somewhat unfair assumption, but the frequency in DS is so low that I can't imagine it would make a difference.

As you can see, DS is literally never used (within the bounds of rounding error). Oddly though, it's only Slow Time/Wither that is used on cooldown. Project should be delayed by nearly a full GCD (on average) in order to avoid force starvation. This scheme gives us TkT every 12 seconds, which is the absolute fastest it can be achieved under average regeneration rates.

The implication here is that Double Strike/Thrash usage only happens when regeneration spikes (due to an unexpected burst of m/r attacks). At all other times, *any* use of Double Strike/Thrash results in a very direct loss of HPS, though potentially a gain in damage/threat.

I'm still not entirely sold on the "thrash-less" rotation. However, it is categorically undeniable that it provides the best survivability. This is a shame, because it implies that a major proc in the Kinetic/Darkness tree is useless. The reason for this seems to be the cost on Project/Shock, which is insanely high. If PA/Energize reduced the cost of Project/Shock (perhaps down to 25-30)

*instead* of finishing the CD, it would suddenly become a dramatic benefit, boosting DPS/TPS without affecting HPS at all. DPS balance could be restored by reducing the (surprisingly high) damage done by Slow Time/Wither. This would reduce AoE threat slightly, but I would argue this would be more balanced, seeing as shadow/assassin AoE threat is already quite a bit higher than what the other two tanks can (sustainably) generate.

**Edit** Here is a slightly different take on the same idea, this time allowing DS to reduce the cooldown on Project/Shock by 3 seconds (it makes very little difference from 1.5):

http://imgur.com/5IivI
The really interesting aspect of this particular expression is what happens when I let the lower bound on z go to 0. Effectively, I can take DS off the GCD:

http://imgur.com/GWJ0O
Thus, if I allow DS to be used off the GCD, the frequency of TkT goes through the roof (once every 5.75 seconds) while Project/Shock is never, ever used without the proc. Forcing at least 1.5 seconds between each DS is all it takes to swing the optimization completely in the other direction, resulting in a complete abandonment of DS in favor of proc-less Project/Shock every 7.5 seconds and Slow Time/Wither exactly on cooldown.