This question does not lend itself to a quick and compact reply; not helpfully anyway. Several factors contribute to the effect.
At least one kind of failure manifests as an incandescent lamp emitting a high, singing tone. If you hear that, first go and fetch the replacement bulb; you know very well that if you switch off the lamp it will never go on again. What has happened is that the filament has already burnt out in a very critical way, such that the spiral winding and its oscillating magnetic field, plus its stiffness and the AC frequency happen to resonate in such a way as to keep the free end of the filament swinging past its matching end while still passing enough current to maintain the light. The most trivial interruption, never mind switching off the light, is generally enough to destroy the effect in that bulb permanently.
There are several other effects, none of which immediately suggests quantum mechanics to me, but I do not intend to attempt to try to prove a negative. Let's just stick to simpleminded mechanics.
The first thing to bear in mind is that the incandescent filament works best at the highest possible temperature. Even a very modest surge added to its working current simply can burn it out, and a slight increase in voltage can shorten its life drastically by increasing the rate of evaporation of metal from the filament. Now, a very important principle is that failure of the filament through evaporation of the metal is a positive-feedback process. As soon as any part of the main heated zone of the filament happens to become thinner than the rest, it's resistance increases and it becomes hotter, increasing the resistance still further. This of course can happen at any time, but there is a strong bias for it to happen to a filament with a weak spot at the time of turning it on. That weak spot usually is quite localised and the cold metal might have a conductivity many tens of times greater than at its operating temperature.
Accordingly there is a surge of current at the time of switching on and the weak spot starts heating faster than the rest by several milliseconds, increasing its ratio of consumption of the current by a large factor. When a lamp fails in that mode you will generally find a sometimes very decorative pattern of condensed metal inside the glass of the bulb.
Obviously there is a strong bias towards this happening at the time of switching on. By turning the current on slowly, one reduces the effect.
Any impurity in the metal at any one point that reduces its conductivity there, has an effect similar to that of a thinning of the filament. The dynamic surge of current at the time of switching it on causes increased stress and heating where the conductivity is at its lowest, in much the same way as suddenly breaking the circuit can cause a spark to jump at the point of the discontinuity.
If one only could add a sufficiently strong negative feedback to the process, causing metal to migrate preferentially towards the hotter spots, one could cause them to strengthen and thereby cool down, but of course that is impossible, isn't it?
Then again, maybe it is not. Suppose one could include in the gas of the incandescent bulb, a trace of a suitable halogen, perhaps iodine. Suppose one could increase the operating temperature to considerably hotter than any temperature that would be practical in a normal glass incandescent bulb, because the filament soon would evaporate. Now the evaporated metal would rapidly react with the volatilised halogen, producing metal halides that typically are more volatile than the metal itself. If we choose a halide that is stable at the operating temperature of the gas in the lamp, but unstable at the temperature of the filament, then metal will be continuously deposited on the filament as fast as it evaporates off. And this would happen at its fastest at the hottest part of the filament.
The hottest parts of course are the weakest parts, and the weakest part therefore would be built up fastest by metal deposition. The glass however, must not be cool enough to condense the metal halides, and that is pretty hot, so it might be a good idea to make the bulb very heatproof, say of silica.Not a bad idea hm?
Not mine, unfortunately!
I sometimes wonder how much experimentation it took to produce the first really practical quartz halogen lamp.There are plenty of other factors, but to keep it short, I'll drop it for now until someone comes up with more detailed questions.
