Switching from 2t8 to 4t9.

2t8: Increases the bonus granted by Eclipse for Starfire and Wrath by 15%.
4t9: Increases the damage done by your Starfire and Wrath spells by 4%.

I've recently decided to switch early to 4t9 from 2t8 (by "early" I mean before I can access ilvl 258 tier), and I just wanted to go over some reasons and logic for that, given that on average I should lose 1-2% dps from the change.

1) 2t8 is not consistent for burst dps.
Most if not all of the math I see on moonkin dps is about average, sustained dps.  I don't have a problem with this as it's the best and most consistent way to thoerycraft, but it is important to realize that it's not all of the story.  There are many situations, especially in hard modes, where burst dps is a very important consideration.  Off the top of my head:
XT's heart.
Stealbreaker
The Saronite Animus on General
Lord Jaraxxus, killing Nether Portals/Infernal Volcanos.
Twin Valks, breaking shields.
Anub p3
(Note: by "burst dps" I mean any dps situation roughly 90 seconds or smaller, where the value of cooldowns will be exaggerated by the short time frame).

The problem with 2t8 in these situations is that it isn't consistent, and I'll use an example from Lord Jaraxxus yesterday.  On the first Nether Portal of the kill I got lucky and it procced with a Lunar Eclipse, so I was chaining starfire crits at 16k each.  On the second, I couldn't get an Eclipse proc and so was only hitting for 7-7.5k, and less often - because of no nature's grace.  So while we killed both in time, my personal contribution dropped by almost 50% simply because of an Eclipse proc.  The fact that I use 2t8 increased that difference.

A second factor in 2t8 is that it simply doesn't scale as well with heroism - you will almost certainly waste a Solar Eclipse proc everytime that cooldown is used, and that means your 2t8 isn't doing you any good.

2) 2t8 increases RNG.
Obviously all classes have RNG (random number generator - in this case referring to how much randomness effects dps), some more then others.  Moonkins with Eclipse - and especially with 2t8 - have more then most.  2t8 is an average dps increase of roughly 6%, which is very good.  The problem is that 6% is only a theoretical number - on any single fight you could see anywhere from 4% to 10% from it, for example.  Possibly more or less - it's hard to say.

Now it's awesome when the stars align and you get a perfect dps parse to top the meters, but it's less awesome when the stars line up, beat you up with baseball bats and you place 15th.  Moonkins are always going to have RNG until and if they change Eclipse, but 2t8 adds even more. 

This is really important now with limited attempt hard modes - before we killed Twin Valks on heroic we had a 0% wipe, and it was close enough that if RNG had been a bit different we might have killed it.  So consistent dps can be really valuable.

3) I have way too much hit.
Ok, this is a personal reason and not one most of you will have, but it's one to consider.  The two pieces of t8 I use are the gloves and the legs, with a combined hit total of 116.  I recently picked up a dagger upgrade and gained 51 hit, which means I am now 87 hit over the cap.  Dropping 2t8 for 4t9 will let me convert some of that extra hit into useful dps stats, and while I might end up socketing a little hit to make the 236 cap, it will be more then worth it.

4) 4t9 will increase my gearscore.
...kidding.  Really.

5) 4t9 allows more flexibility in rotation. 
Another weakness of theorycrafting is that it generally only considers single target situations.  There are a fair amount of situations where this isn't the case - Yogg-Saron p2 being the biggest one right now.  This is a fight where dotting multiple corrupters plus the crusher isn't padding the meters but is a real necessity for the hard modes - that stuff needs to die consistently for p3 to be smooth.

The problem with 2t8 here is that it increases the opportunity cost of ignoring Eclipse - because so much of your dps is tied into that one ability, ignoring it to multi-target DoT really hurts.  It's still a dps gain, but it's less of one.  4t9 isn't tied to Eclipse, so it becomes more worthwhile to DoT up a bunch of things and just make use of Eclipse where you can.

Another benefit is that by reducing the benefit of Eclipse, you can reapply DoTs slightly more often and still see a dps gain - instead of being 6+ seconds, you instead want to reapply at 4+ for IS, and 3+ for Moonfire.  That leads to a bit more consistent rotation and damage, because more of your dps is coming from those DoTs instead of more RNG-dependent spells.

Those are the main considerations I made before deciding this.  I consider those reasons pretty valid, but of course your mileage may vary.