New LED fixtures can reach quite more than 100lm/w. For example this zumtobel fixture reachs 136lm/w @ 4000k with a cri >80 included the ballast losses. I've saw a shop completely lighted by these fixtures and the light quality is exactly the same as 840 trisphosphor tubes.

Is the degradation caused by the metal reaction also part of why that some MH bulbs fail explosively? I've also heard of the sodium in LPS lamps reacting with the electrodes and ending up eating thru the glass.

@Kappa7: LEDs suffer from light quality and brightness diminishment. The color of the LED light can turn less appealing due to the plastic casing discoloration or the LED die simply wearing out. They also suffer from lumen depreciation.

The LED lumen depression is strongly linked to the operating temperature, so if you keep them cool, they will last way longer.
Otherwise their problems are the early failures, mainly loose bondwires.

For the MH explosion and darkening: Yes, it plays the role, although not directly. It traps the light, so causes more power ending up by heating up the arctube. That means higher operating pressure, that means higher stress, plus larger arc voltage. And if you have really constant current ballast (typical example is the CWA), the larger arc voltage means higher power input, heating the arctube even further. Until it reaches the point, when the arctube can not hold the pressure anymore.
If you have a ballast detecting this and either limiting the power or shutting down (don't know about any other than electronic with such features), the chance for the explosion is greatly reduced. But not eliminated completely (the fatigue cracks and ballast failures are the other frequent causes of explosions), so the use of such ballast still does not replace the required blast protection (protected lamp, tough fixture,...).

@Kappa7: LEDs suffer from light quality and brightness diminishment. The color of the LED light can turn less appealing due to the plastic casing discoloration or the LED die simply wearing out. They also suffer from lumen depreciation.

In case of these new zumtobel fixtures the The lumen maintenance is 90% @ 50000hr so it's really good(probably the LEDs are run at low power and kept quite cool) and not worst than long life fluorescent lamps. The plastic casing discoloration should be low if the LEDs are kept sufficiently cool, because the LEDs doesn't emit UV, the heat is the only cause that can produce a discoloration. Other than that they have also a 5 years warranty, so I hope that they are quite reliable, but only time can tell this.
BTW I'm not  a fan of LEDs, but I have to recognize the enormous advancement that they are doing in these last couple of years. "Sadly" now it seems that they have definitively won the efficiency battle with fluorescent lamps.

That is how they keep them cool in some lanterns. I doubt the fan will last as long as the LEDs or driver, in the dusty and spider riddled environment of the space in the ceiling

http://www.lighting-gallery.net/gallery/displayimage.php?album=2438&pos=14&pid=75662

There was also talk of LED lanterns cooled by peltier units. I don't know how they plan to get rid of the heat generated by the other side. That would be back at square 1 with the overheating problem (assuming that they use just a heatsink to draw the heat out.)   Also, newer LED lamps are built fairly cheaply so they're prone to overheat.

I don't think the use of Peltiers makes that much sense in general lighting: For a 1W transferred they consume extra 2..3W. And as the 1W is 80% of the lamp power input, it means 3..4x higher power consumption of the fixture, where is the efficacy then. And when left in the passive mode (I would expect the active cooling would be used only in the peak ambient temperature moments, so very limited time; just for the efficacy reason), the Peltiers form extra heat resistance.

At that rate, they'd just be better off using HID instead.

I noticed that no manufacturer makes a decent high colour temperature MH, is InI that corrosive, or am I missing something here?

I think that it is what Medved said about certain salts being corrosive to the arctube, so the lamp will have a shorter life and have a better chance of catastrophic failure.

Mainly the sodium, but as well many other metals from the left side of the Periodic table as well, are highly corrosive towards quartz, it is once the salts dissociate in the arc body and then some of the metal ions find their way towards the quartz wall. In fact they are responsible for it's blackening on MH. To prevent that, the lamp design tends to keep all things around the arctube on a positive potential towardss the arctube content, so the electric field repels the ions away from the quartz wall (only ions are such problematic, once it recombines with the halogen back to the salt form, it is again quite inert towards the quartz).

For the ceramic MH, the mechanism is different: The metals are inert towards the alumina, but the liquid salts alone tends to dissolve it slowly. Therefore the lamp design always tend to keep the pool on place, where the arctube wall could easily be made thicker, without blocking the light output. And that is, where the ball shape arctubes have their benefit over the classic cylindrical shape: The walls where the light should pass through, are closer to the arc, so operate way hotter, so prevent the liquid salts from residing there - they are distilled to some colder place behind the electrodes, where the arctube is made thicker, so some dissolving won't compromise it that quickly.

You say that the sodium is the most corrosive, then how come the HPS bulbs don't show degradation or explode at the end like MH bulbs? Is it since HPS bulbs use alumina arc tubes instead of quarts or ceramic or borosilicate.

Sodium is corrosive towards the quartz, that is the reason, why HPS don't use quartz (nor glass) for the arctube, but sintered alumina - towards that the sodium is rather inert.

LPS use some sort of hard glass for the arctube, but there is special protection layer covering the inner side separating the sodium from the glass itself. That is quite complicated and expensive technology, so a reason, why the LPS were ever made by just two makers on the planet (all the others are or were made in a commercial scale by dozens of makers) and the reason for their rather high cost.

Is that also why some MH lamps are strictly restricted to one burn position or the other?