Because analysing DPS by looking at formulas is a really tough job when taking everything into account (GCD, DoT-timing, procs, resists, GoN-procs etc.), I developed a stand-alone boss simulator for moonkins, to generate a huge sampleset of empirical data without the struggle to actually fight a boss.
This program takes any possible mechanic into account (excluding partial resists) and gives the possibility to simulate hundrets of hours of nuking within one minute. Mana is not yet implemented, but also not important for what I was looking for. The goal of the spreadsheet I made was to look at the sheer dmg potential of different talents and builds.

The testing was done without taking any buffs into account, except for the 5% crit and 3% haste from the moonkins impr. aura. E&M is also left out even when skilled, because it is a flat modifier that has no impact on percentages when comparing the builds. (Haste has because of the GCD.)

I would be VERY gratefull if anyone could post this in the US betaforums, mostly because it shows the underwhelming performance of Eclipse and the impact of (aparantly mandatory) manareg-talents versus more dps talents in hard numbers!

The Builds and any other presumptions that are important are in the excel sheet, as well as my remarks of which I will post some as a teaser :
- Starfirespam is not far behind Wrathspam (mostly because of the Glyphs).                                                            
- Builds with 3/3 Eclipse and perfect use of it are roughly in a range of 2-3% DPS increase to pure Wrathspam using a just slightly different build. IMHO not worth the struggle you have keeping track of the buff all the time, thus having less eyes on the fight itself.

The spreadsheet can be downloaded here: http://rapidshare.com/files/145793519/MK_builds.xls.html (Yes, it is clean and anything. Not using fancy makros anyway.)

I copied your post over, as well as provided a link to the original.
http://forums.worldofwarcraft.com/thread.html?topicId=9879057813&postId=100422505482&sid=2000#181

Thanks for taking the time to do this sort of analysis.

I realized I didn't state your name for credit in the post, that has now been fixed with a thanks to Madrocks.

i dont understand how this would be just slightly more dps? all ya need is 2 crits with starfire... and then all of your wraths hit for 10% more damage for 30 sec.. then ya do it again..... and with no more mana pot chugging, you wont be spamming wrath all the time or you will be putting the OOM in Boomkin.

seems to me a chance to regen some mana and get a 10% damage buff.. am i missing something?

I have to beg again for someone posting the text below to the off. US beta forums, into this thread please. Cookies are out for now, but I made some great chillie con carne today which I'd be willing to share =) Thanks in advance!

8<-------------------(snip)-------------------->8

The recent changes sound really nice, except that I had bad suspicions about the Eclipse change from the beginning. I changed my simulator to adapt to the changes and ran some new tests. The outcomes are in short:

- The spirit conversion provides roughly 100 DPS increase with all-Naxx25 leather gear. Very welcome, because it'll scale with gear.
- The changes to Nature's Splendor now make this better than impr. Moonfire, despite the better chance to gain some mana with imp. MF. (It is about two times more effective expressed in percentages.) This is also true if the MF-Glyphs are equipped.
- Eclipse, finally, is completely worthless. I can't find words for HOW worthless it is now. It now gives a whopping 0.1%(!!!) increase in DPS if used PERFECTLY. The best way to use it, is casting Wrath until SF-Eclipse is up, casting SF for that time and then switch back to Wrath. Perfect use means now, that you instantly recognize when the buff procced, which is impossible because you're already half through casting the next Wrath when the former hits and procs.
Anything you do other than the rotation I just explained will LOWER your DPS to be under that of pure Wrath spam and keeping the DoTs up. If you go that far to invert the rotation (what is what most people will belive would be the better way to use it anyways), you lose up to 3% DPS, even if you time perfectly. This means that you actually are doing nothing wrong, and spent 3 talent points to lower your DPS. Add the amount of awareness you have to direct to buff-and-CD-watching, and therefore away from the actual fight, and you finally can say that this talent will always make you play worse. Always. Get rid of it. This isn't even remotely funny anymore.

The whole concept of Eclipse needs to be changed, and there are hunderts of really nice and non-overpowered ideas in all official forums including this one. But IF you change Eclipse to something worthwhile, please remember that you your selves have announced to take care of the bloat in balance. Right now it becomes very hard to skill properly, and with one talent more it'll become downright impossible.

For reference to my former testing, and some words on what "my simulator" means, please look here (2nd posting on that page): http://forums.worldofwarcraft.com/thread.html?topicId=9879057813&sid=2000&pageNo=10

I did not investigate the impact of IFF for now, but I did some testing how much percent of DPS you lose when recasting it every 40sec some time ago. It cost about 5% of our total DPS. So if we assume a SPriest in our raid (which is quite likely), the 3% crit will not make up for the loss you have casting it. With gear getting better (=more hit) the hitbuff will additionaly be obsolete in later instances, leaving the 3% crit the only effect in any circumstances when raiding gets to a point where minmaxing becomes much more important.
This change, combined with the impossibilty to take this talent thanks to bloat, won't make it desireable anytime.

I strongly belive that the crit is linked to the improved FF. It would not even be hard do discriminate between a talented and untalented debuff. If it gives X% hit, make it give X% crit as well, and you're done. Even if what you suspect is true, the 3 points can be spent on more effective talents. 3% crit unfortunatly do not always give 3% DPS, if not hit-capped never. Better keep 2 points in MSS, skill E&M 5/5 or get Celestial Focus along imp. IS. All these would be better places for these 3 points.

Well, for one, recount doesn't look too convincing in this version - glancing casts?

The problem with "live" testing is, that the RNG'ness renders any sampleset below some several thousand casts more or less useless. When simulating with my program, I still get variances of about 10 DPS even after simulating 50 hours of constant and perfect nuke. It is too, not so good to incorparate trinkets and idols when aiming to rate the effectiveness of talents.

I however can imagine, that when popping trinkets with a max. 30s runtime while the former form of Eclipse was up could do the trick to change the SF->Eclpised-Wrath rotation to be the better one, but I wonder what would be if you used the trinket outside of Eclipse to buff the un-eclipsed cast. Especially now, where every trinket will run longer than Eclipse lasts.

But the main problem is still the very limited sampleset from live testing. To make reliable observations, you almost have only just two choices: Make formulas, or if like in this case that can get really ugly and close to impossible, get a much greater sampleset.

I just wanted to remind you, that my (now cold...) chillie is waiting for someone who could post the feedback above (the one with a blue header) to the US beta forums. (Yes, it is kinda important to me )