Here's a possibility.
Instant start ballasts effectively start the lamp "cold cathode" style - with a high voltage pulse.? This is particularly hard on the emitter coating on the electrodes as there's no thermionic emission taking place due to electrode heating (the preheat fascility is unused).? Needless to say, this means that a lamp operated on a ballast of this type will likely fail prematurely if it's switched regularly.
Now, the reason for the clear ends (please note - this is a guess!).? Due to the fact that starting's so hard on the electrodes here, these lamps tend to fail due to exhaustion of the emitter coating on the electrodes.? As the emmitter is sputtered away, the area which can emit electrons to the arc will decrease.? Obviously, this area will run hotter and hotter as its area decreases, and the rate of sputtering will increase.? Eventually, all the emitter material will be exhausted - at this point, one of two things will happen depending on the type of ballast.? 1. The lamp will go out, and that's it.? Some ballasts detect the end of lamp life and shut it down.? Or 2, the lamp will continue to light - as the ballast can provide enough voltage to sputter electrons right off the surface of the electrode - in which case you'll get a lamp that's still running - albeit somewhat erratically most likely, in this case substantually higher temperatures than usual will be present - possibly even high enough to cause the glass near the electrode to crack (I have seen this happen on two lamps - both compact fluorescents).?
There is another possible way that can happen - which is somewhat more unusual (i.e. less than 1% of lamps failing in that manner) where the electrode breaks, and remains attached to the terminal at one end - but the other end comes into contact with the ring of sputtered electrode material on the inner surface of the tube - this then behaves like an electrode, and will maintain the discharge for a time, until the resultant temperature rise causes the tube to fracture or the elecrode loses contact with that area.? I've never actually heard of this happening, but I've heard it theorised.
IMT Lighting Mention this on their webpage too - and they seem to know what they're talking about - When you're talking about equipment in that sort of setting, you don't take chances.
If something is causing the tubes you have to rupture, the resulting inrush of air could possibly blow the phosphor (which is of course a powder) off the surface near there.
All just theories, but I'm bored!? So having something to think about it good!
As an edit, just dug out the GE Biax lamp which I had fail due to something similar to my first theory.
One tube end has blackened very severely (for nearly an inch up the tube), and has cracked right next to one electrode. I was present when this lamp failed. It started flickering like a mad thing at that end (other electrode seemed to be doing fine still), for a few minutes, a very intense blueish white glow present there - before it went *tink* as the tube cracked, and a then it flickered for a half second or so and went out.
