Granted, once 3.2 is released, this becomes a bit of a non-factor but I wanted to address the "soft-cap" of 400 haste for rotations. I will open by making the statement that for a Lunar Eclipse, the 400 haste cap is asinine. That's right. It's complete fodder. By definition, the Lunar Eclipse has a 60% chance to proc on a Wrath crit, and while I realize sometimes we have a dry run with many Wrath crits and no Eclipse proc, that is often only our memory latching to the bad instead of the average.

Ok, so why is the haste cap ridiculous for a Lunar cycle? These events are independent (i.e. one Wrath crit does not impact when the next Wrath will crit). Instead of looking at Eclipse as a 60% chance to proc on crit, let's look at it as a 40% chance to not proc. Let us then look at the successive chances that Wrath will not proc Eclipse.
As we can see, having 5 independent crits with no proc is a 1% chance. Now, what does all of this have to do with the 400 haste soft cap? Well, the haste soft cap includes the Nature's Grace haste bonus. If you're not under the influence of Nature's Grace, you can afford more haste.

The current logic states that any rotation that contains Wrath should have a soft cap of 400 haste and I'm afraid I've just seen too many erroneous posts asserting the anything over 400 haste is a waste. It's not. Any rotation that contains many Wrath casts, including crits, with the intention of continually casting Wrath after the crits (read Solar cycle) should aim for a haste rating of around 400. Any rotation that plans to switch spells after a Wrath crit is more than okay with a haste value over 400.

I've been puzzling over how to represent this mathematically since it's very tough to model the probability of some chance encounter when I more or less know the previous events. The problem we encounter is that probabilities are not static once the sequence is in motion. Once I've gotten my first crit and the Eclipse did not proc, I reset the entire sequence. Due to the fluid nature of this problem, I have decided to make a table of likely sequences and attribute each sequence a percentage. That way I could see just how likely I was to be casting under the influence of Nature's Grace. I assumed a worst case scenario (which is that all the crits have 3 seconds of Nature's Grace associated with it). Below are the weighted values of the Nature's Grace time through n = 5.
As I sum this, I find that I get a value of 3.198 seconds of average uptime on Nature's Grace given my worst case scenario of non-overlapping crits. I would be working with for n=1 through infinity. I treated this like a Power series and came to the conclusion that it summed to 3.333 repeating, of course. Now, I did this over the course of several 5 minute spurts throughout a few weeks, so I could have fractured logic (in which case, tell me!!) but I'm forced to come to one conclusion here.

TL:DR

The results seem to indicate that I have an average of 3.3333 seconds of uptime on my worst case scenario in a Lunar precasting phase. That means that for any haste over 400, we clip the GCD for an average of 3.3 seconds. I really don't feel like this is such a bad thing because I showed earlier that the average Lunar cycle takes around 35 seconds. The rest of the rotation (32 seconds) is spent without clipping the GCD. That's 91.4% of the rotation that benefits from haste over 400.

The haste soft cap is exactly that: Soft. Especially for a basic Lunar Rotation. Naturally, math gets more complicated when Starfall is accounted for, although then we're talking 67% of 91.4% or 61.2% of the rotation. I say this because Lunar cycles are using Starfall at most every third Eclipse. If you plan on running a Lunar Eclipse (until 3.2), haste is still good for you. Don't be afraid to go over 400 haste!