I just did a new lighting installation which, for the first time ever ive installed electronic MH ballasts that can operate an array of lamp wattages (150w, 250w and 400w) controlled by a selector switch on the side of the ballast. What I found also interesting is that if a 250w or 400w lamp is installed, it can be dimmed by switching the wattage control to the next lower selection (eg, 250w lamp running on 150w, 400w lamp running on 250w. The manufacturer however doesn't recommend running a 400w lamp on the 150w setting. What do you guys think of this? I know significantly underdriving MH lamps is damaging to the lamp, but thats on a magnetic ballast. Is it a different case on an electronic ballast?

Btw I tested a 400w standard 4000K PSMH lamp on one of these ballasts switched to 250w, the color of the lamp actually wasnt as bad (greenish) as I thought it would be due to only partial halide evaporation. Kill-a-watt reads 284w

The reasons why the deep dimming is not recommended are lamp performance (color, efficacy) and the fact, the lamp operation is not reliable (unstable arc, it tend to extinguish), mainly when lowering the power level (the transition should be done smoothly or with small steps and slow, so the arc tube temperature could follow, so you do not operate hot tube on low power).
So such installation would be not much usable.
The first aspect does not depend on the ballast, it is the same for magnetic, as well as electronic.
The electronic is generally capable to maintain the arc way more reliably, so the second aspect would be less of a problem with the electronic. The problem with the electronic may come from it's lamp diagnostic features (the triggers for the lamp EOL shut down), they may activate.
So if the ballast is rated for that use (lamp dimming via the power selectors), it will work well (but only within the rated dimming range). But if the ballast is designed to just run the selected wattage lamp, or when exceeding the range, the dimming attempts may end up by the ballast shutting down the good lamp, so an unreliable installation.

And what is important with any dimming system, the lamps shall run at full power the first 15 minutes after EACH ignition (power ON and even restrike) and only then the power could be reduced. That is important to first ensure the lamp warms up completely and so operates in a correct mode and (mainly in MH) the inner chemistry has the chance to recover from the ignition damage (halogen cycle clean out the arctube wall from the sputtering,...).

Wouldn't the cathodes sputter if it was underdriven, despite being allowed to run up all the way at full power first?

Not significantly different than when all the time on the rated power, they are less heated, but less loaded as well. The main problems are with the halide salts condensation and cleanup after the start (the main sputter happens during the start anyway)

Well I ended up contacting the ballast company and they claim that "this ballast keeps the lamp perfectly stable at the reduced wattage and in no way affects the lamp, or lamp life negatively, even if burned at reduced power the entire life of the lamp, it will still last its rated life." now whether thats true or just marketing BS is anyone's guess and only time will tell. However, I do know that when a 400w Na-Sc lamp is operated base up and running on the reduced 250w setting the color shift is very minimal, in fact only slightly cooler than a normal 250w MH lamp, no green hue at all. Also the arc stays perfectly in one spot and doesn't swirl or move about, no flicker either. The ballast also runs the lamp at line frequency square wave, which is unusual for an electronic ballast

That means the ballast is designed for such operation, so I would reallynot expect any problems.

It is true, when the electronic delivers really the constant, programmed, power to the lamp (unlike the CWA, which delivers constant curent), the low pressure "deadlocks" are avoided (the ballast respond to the lower arc voltage by elevated current, so the arctube heats up to the final temperature anyway).

However (I do not know if that apply to the the NaSc as well), but the European rare earth chemistries tend to evaporate only some of the components at reduced power (the individual components difer in the boiling point), so the fill tend to separate. The occassional full power operation evaporates all of the fill, so mix it back up.

So I would read the manufacturer's claim "...all it's life..." as "... all it's life*..." with the footnote "* of course except the first 15 minutes after ignition, as recommended by lamp makers"
Anyway letting the lamp operate at full power the first 15 minutes after ignition won't hurt at all, so I would include some timer for that...
But generally expect worse performance (efficacy, color,...) when operating 400W lamp at 250W than when using directly the 250W lamp, so the dimer system has it's place really only when you need the output to be variable.

Yes, the reason these dimming ballasts were utilized was that the dimming control was needed during a few 2-hour "breaks" during the day where the facility was not occupied, yet lighting was still needed as some of the staff was still there working during those 6 off-peak hours. The ballast's dimming feature uses radio frequency and with the push of a button all 30 fixtures dim to 250w, then when full work production resumes, press the "400w button" the lamps ramp up to full 400w. During night only a single circuit (6 fixtures) remain on the 250w setting, and all others off. The company also gets a decent energy-efficient upgrade tax write off as well. After the install there was 2 ballasts leftover, which I kept for experimenting. Interestingly enough these ballasts will not run mercury lamps, after ignition the ballast shuts down in about 3-4 seconds, and will not attempt to restrike until power is cycled on and off. HPS lamps (400w) will run up just fine to full intensity but as soon as you switch to 250w setting the arc extinguishes. Only metal halide lamps work with these dimmable ballasts.

The mercury has different startup characteristic, so the ballast "think" the lamp is EOL..

Or it could be just an artificial measure to discourage the use of MV's...

Otherwise I think that sounds like good concept. With the night operation, I would make the group of "nightly" operating fixtures roling, so each night it will be another group. One reason is even wear (so there won't be significant differences among the lamps), second to regularly cycle all lamps, so if the lamp would be going to explode, it will do so during warmup (because at that time the stresses in the lamp structures are the greatest), when there is no yet as high pressure, so the eventual explosion won't be as violent.

Took one apart and thought ide post a pic of the electronics. As you can see lots packed in there, in fact there are 5 "sub" pcb boards mounted vertically off the main board, two more in the 1" gap on the opposite side of the main board as well. Interesting thing is one of the sub-boards has "400w-600w -DRIVE" printed on it, yet this ballast only accepts a max of 400w lamps, there is no 600w setting. Not to mention a 600w MH lamp doesnt even exist as far as I know.

The small boards are usually controller parts, the power stages are then on the main board.
It allows the more complex circuits to be independent, so could be stocked regardless of what output power would be used for.
The "400-600W" label suggest this platform is likely used for some special lamps as well...

But generally expect worse performance (efficacy, color,...) when operating 400W lamp at 250W than when using directly the 250W lamp, so the dimer system has it's place really only when you need the output to be variable.

In the streetlighting applications I have seen the dimmer system is indeed used for a variable output. But how: Until 10 PM the lights are run in the described way, at an estimated 60 percent of nominal power. Then for the rest of the night the system goes down to 30 percent or thereabouts, perceived as a glow with a distinctive colour shift towards deep yellow or green, respectively.

At least the latter looks so unhealthy that it immediately made me wonder if this will really not affect lamp life. If it does my Schadenfreude will be with them, although one has also to wonder if perhaps they are rather victims of an approach to transfer the responsibility for violating EN 13201 to the operator of the installation, wanting to save money while not really knowing what they are doing.

30% rated power on an HID lamp is very low and sustained operation will likely produce some negative affects. Those dimmable ballasts I installed wont even operate a 400w on the 150w setting, it just cuts out about 20 seconds into warmup and wont restrike the lamp again until power is cycled on and off or a higher wattage setting is selected (eg. 250w, 400w). Electronic ballasts being more sophisticated can more properly dim a lamp while minimizing the spectral shift and reduction of lamp life. Those 400w lamps running on the 250w setting have an almost indistinguishable color temp from the same lamp on full power (400w). How they get temp up to vaporize the halide salts on such low power im not sure of. 

The color shift is directly related to the fill component ratio, so to the temperature. If a 400W lamp does not shift on 250W, that is a feature only of the lamp and not of the ballast.
Power reduction down to 30% usually does not as much affect the lamp life, but mainly it's performance (color, efficacy). When that is of no problem (streetlight at really quiet hours,...), there is no problem with that, it still saves some extra energy.
The compliance to standards like  EN 13201 or so is not mater of just a lamps, but always the whole installation. And for that the operator bear most of his responsibility anyway (it starts with overall design, pole and lantern selection, the use of an eventual dimming is quite minor part of it, ends in how it is operated and maintained), so when all is done well, there is no problem. And if it is done wrong, I don't think the dimmer would change anything significant there...