I have not done than as a warlock, obviously, however I have done them and I do know the things that have caused us to end in a wipe.

No matter how fast I move, when I have a Solar Eclipse running (thus both Sunfire and Insect Swarm dealing very high damage, and I keep both dragons DoTed constantly) and Starfall ticking off all at the same time and I get Engulfing Magic -- people will die. The damage that I am able to deal to a raid via merely uncontrollable DoT ticks is rather absurd, and our healers simply cannot keep up with that.

For this reason, I watch my Engulfing Magic timers, I won't re-apply my DoTs if it's coming up within around 5 seconds, I won't use Starfall either, and if I do get it, I remove Starfall immediately as I run out. All of these things reduce my personal DPS; yet it saves my raid quite a significant amount of damage.

When dealing with a mechanic such as this, you do have to think about it logically and not in terms of personal damage output. Is using the Glyph of Soul Swap a DPS increase? Yes. Should it be used? Potentially no. The gain here is merely one of personal standing on the meters; which, although I like to shoot for the highest possible position, I regard as secondary to actually downing a boss.

Is the situation going to be that your raid is in grave danger should you use GSS? Probably not, but the potential exists. You have additional DoTs running, you have no method of controlling said DoTs, and you cannot control additional sources of raid damage.

If myself, as a balance druid, and an affliction warlock both get Engulfing Magic -- it's going to strain our healers. Same goes for any combination of heavy DoT/HoT specs. Toss in a Fabulous Flames or Meteorite landing within the same short time span, and there's a very large potential for people to die.

The damage increase from using GSS is significant enough to warrant using it, I concede that point; however almost never is damage actually going to be the reason that players wipe on this encounter. That is why I argue that choosing to lower your damage potential slightly in favor of increasing survivability chances is the better option. Were this a DPS race encounter where the damage you dealt had a more significant impact on whether or not your raid wiped; then I would lean more towards increasing damage. That isn't the case with this encounter.

I would also not agree with your assertion on Halfus. Clearly there are guilds that allow multiple players to sit on Halfus for the entire encounter, otherwise we wouldn't see the damage parses that we do from many of them. Given that the evidence suggests that a destruction warlock would yield higher DPS via using BoH on an add and DPSing Halfus; I do not see how healer mana even comes into the issue.

Healer mana is constrained by the encounter length as a whole, not merely the first minute; Halfus will have to die eventually, and if using this method yields a player their highest DPS potential, then why not? There are two methods for doing Halfus: releasing the drakes one at a time, or two upfront, and then one.

When you do two upfront, you should be using Bloodlust right at the start. In a 25-man raid, having one person putting out more DPS on Halfus than on the add won't make that significant of a difference in how quickly they die. I could see an arguement for using BoH on one add while killing another add at the start -- especially in a 10-man group -- yet this contradicts your previous statement.

It is okay to potentially increase raid damage in order to increase your personal DPS on T/V, but it isn't okay to increase personal DPS at the potential cost of additional raid damage on Halfus. Why? The drakes will die either way, your DPS will increase; so why shouldn't it be done?

It again comes down to risk/reward. How much longer with an add live if you use BoH on the add and DPS Halfus instead? How much will that increase tank/raid damage? It's all a balancing act; I contend that the scales in this encounter tip towards increasing damage because Halfus -is- a DPS race of an encounter whereas T/V is not.