Google translate link (Original hebrew link) .
Article about water cooled LED lamps.
They calls the fluorescent lighting "Neon lighting".
They even goes further to say that LED lighting have more natural lighting that fluorescent ("Neon") lighting.

First of all, I'm not a big fan of LEDs.

Disadvantages:
1. If you need a high output, the LEDs must have a huge cooler. Otherwise, their life is radically shortened.
2. LEDs are too directional. It can be advantageous in some cases but it is usually a disadvantage in case of indoor lighting. An additional optics may be required.
3. Glaring. The light has to be diffused to prevent it. Diffusing descreases the efficiency.
4. Multiple shadows. Usually, the LED sources consist of multiple chips which means the light sources in different positions.
5. Cheap chips have a low efficiency. As far as I know, some COBs do only 70lm/W which is worse than many fluorescents.
6. The lamps are usually rather ugly. (That's my personal opinion.)
7. Quality power chips are expensive.
8. The durability of LEDs is uncertain. I've seen a few of them dead after 3 years in the low-end street lighting. I'm sure that manufacturers employ the planned obsolescence here. (See the "light bulb conspiracy" document.) They do not want to close their factories in the future.

Advantages:
1. The perfect application for LEDs is in torches of any kind. It's a small and directional source which is resistant to vibrations and shocks.
2. No warming up. 100% output instantly.
3. Quality chips have good colour rendering. If you use a spectrograph, the spectrum is almost perfectly continuous.

Conclusion:
In my personal opinion, at this time, there is no rush to replace fluorescents in the home/office lighting and HIDs in the street/factory lighting. The savings and the performance boost are at least uncertain.

Merc: The spectrum of white LED, consist of a narrow blue band (Blue LED chip) and a wide band from the red to the green with a peak in the yellow (Ce: YAG phosphor). It isn't possible to produce a continuous spectrum with a blue chip and a yellow phosphor.
The CRI have nothing to do with the pleasantness and the comfort of the light.
Most daylight LEDs have CRI of Ra8=80, and still causes me a headache, looks harash blue-gray and causes a blue glare effects compared to halophosphors fluorescents. Even daylight halophosphors fluorescent lamps (Ra8>=70) are more pleasing to me than daylight LEDs. SBMV lamps also more pleasing to me, despite of their Ra8=65.
There are people that the light of CFLs causes them harash as well, despite of their Ra8=85.

Merc: The spectrum of white LED, consist of a narrow blue band (Blue LED chip) and a wide band from the red to the green with a peak in the yellow (Ce: YAG phosphor). It isn't possible to produce a continuous spectrum with a blue chip and a yellow phosphor.

I agree. That "almost perfectly continuous" was exaggerated a bit. There is an article http://protonsforbreakfast.wordpress.com/2011/03/25/light/ with downloadable data from the spectrograph. The LED spectrum looks really better than the spectrum of CFLs.
Another comparison: http://housecraft.ca/2012/09/30/the-diy-decorator%E2%80%99s-eco-friendly-lighting-dilemma/.

You are right with the LED discomfort for the eyes. That's also my experience.

The "harsh" feeling come not only from the cold color, but mainly from the fact, the LED's are rather sharp, pointy light sources. Even when there are many of them, the "thousend sharp shades" effect they made is quite irritating. The larges reason for these products is, the LED lighting is "cool", so makers want to stress, the light is of the "sexy LED" type...
Other aspect is the very poor matching with fixture optics: Fixtures are usually designed for rather compact omnidirectional incandescents. Nor CFL's, nor most of the LED match that. The result is very uneven illumination, dark ceiling, overlit spots and so on.

And for the color: The LED's are in fact just two band devices: One peak in blue and then a wider band in yellow. Even they may somehow trick the spectrum analysis to get the results of Ra of 80+ and CCT of 2700, it still look unnatural. Moreover even when most products claim >80 CRI, when measured, it goes barely to ~60..70, mainly on missing red and blue-green area, so it look really strange yellow. And for a warm 2700K tone this look very terribly (a 6500K CRI60 look way better than 2700K with just a CRI 70).
The "tri-phosphor" CFL's have distinct lines, but they are four (not just two) and rather evenly spaced over the visible spectrum...

And mainly with COB's (but slightly with other format as well) is another frequent problem: Although the COB present itself as a diffuse light source, quite well diffused is just the converted part of the spectrum. But the blue primary LED light come from very few sharp points. So when this is not properly mixed with an additional diffuser, it does create strange feeling as well.

And a separate aspect is selecting an appropriate light color for given application and light conditions: Way too frequently I see very strong 2700K light sources, what mean when used, the intense illumination at 2700K is rather strange yellowish. While for a strong illumination you need really way higher CCT, 4000K is a minimum.
On the other hand I see weak "bed-lights" with 7000K LED's. Again no wonder they look bluish-gray, that low levels are asking for low CCT, no more than 3000K.
But these are not faults of the LED's, nor fluorescent, nor CFL, but the one, who have chosen such inadequate color tones for those applications. Doing this correctly really makes the main difference, even when the lamps aloe are on the worse side (CRI is low,...)

I've noticed this too with the Luxeon Star LEDs .. and also "white" (=>6000°K) LEDs .. the light distribution does not match the spectral distribution, giving coloured bands towards and beyond the edges of the main beam. Narrow beam angle sources (not sure if I actually consider them as "lamps" as such, as the technology was never intended as a light source, more as a rectifier and producing light for indication purposes as a by-product) produce rings with alternating colours beyond the main focus. Whilst this would be a nice effect in crystal chandeliers such as the one I have in my front room, as it would accentuate the prismatic properties of the lead-glass, it would actually be a problem in any other application.

I don't mind low CRI and actually love the light produced by clear mercury lamps even if it did initially take a bit of getting used to (when I first tried this I was very conscious of the strong yellow, green, blue and violet bands) .. I now find that the colour effects are very pleasing and restful to the eyes, plus they don't interfere with other light sources such as the TV or computer monitor .. you can see well and there is no glare whatsoever.

But having unequal spacial spectral distribution is going to cause problems, much as using a strongly colour-biased light source such as HPS which in itself makes TV viewing difficult due to the light source washing out the picture and altering the eye colour response. This is by far the biggest problem for me with indoor lighting.

Whatever other light sources I've experimented with I've always gone back to clear mercury in short order!


BG