Here's a GE C50 lamp flickering and flashing like crazy, and then finally losing vacuum!
http://www.youtube.com/watch?v=P37dX3BlE9g

Operation on a slimline ballast or an instant start electronic ballast without any protection mechanisms. They have the ability to continue to operate the on cold cathodes when the electrodes are worn out, overheating the lamps end and leads to cracking in the end seals, eventually leads to vaccum loss.

dor123, it's different, it melts the electrode leads to the glass, where it continues arcing into the hot glass, which causes it to fail with vacuum loss!

I got this to happen on a 20W quick start ballast (similar to rapid start) once! There was a gap in the phosphor so I could see inside the tube, the cathode filament had disintegrated and left only the support wires. I left it on rectifying for a few minutes and then it suddenly came back up to full brightness. I looked down the end of the tube to see what was going on and the filament support wire was glowing white hot and a piece of molten metal dripped down on the glass and cracked it.

I didn't think standard 240V mains would be enough to cause a 20W tube to run in cold cathode mode!

Depending of the ballast the lamp is run on, the lamp's failed electrode may search to provide thermionic emission at any cost, including the cost of its own destruction. That is (I think) why the electrode is hot like (...) when failing. The hotter filament generate what's called the "Edison effect" in order to compensate the lack of emissive substance necessary for thermionic emission at the proper temperature. Running a "hot cathode" white hot is obviously highly destructive! If the filament can run that way for a while, its supports can just not stand that condition at all, so they melt right away. The glass stem cannot either, hence the vacuum loss

Nice video BTW! Oh and which ballast the /C50 tube was run on?

Vince, this was run on an F32T8 ballast.

Vince, this was run on an F32T8 ballast.

He meant the ballast concept and exact characteristic (e.g. protected vs unprotected, how the current depend on elevated arc voltage, how long might it might in the high voltage mode,...), not the exact lamp type it is designed for.
For any lamp you might have one ballast doing these and some another not doing these effects (and another doing other crazy things with such lamp).

As i do not like uncontrolled fires, i more like the second ballast behavior...

I am still looking for a ballast that can run lamps up to the vacuum loss. The only F32T8 electronic ballast I have do the EOL shutoff thing, BORING! I can't even wire a jumper on the broken cathode to fool it, doesn't work! 

@TiCoune66 all you need is a F32T8 Instant Start ballast - the type with only one wire going to each end of the lamp ... works with most brands i've tried (GE being the exception)

LOL fun way to 'finish off' a dead lamp

@TiCoune66: With such ballast you would be asking for socket melting (at least they would be very crispy after that), very short ballast life and a fire.
I think this s the reason, together with stricter safety requirement, why you do not find such ballasts (one wire per lamp end, connecting both filament terminals together) in Europe

@xmaslightguy: I think the danger (see above) makes these ballasts disappearing even from US market:
I red already few years ago in some official document(or it was Code or DOE/EPA specification) and find there a requirement, then "ballasts SHOULD prevent socket from overheating due to faulty lamp behavior" and as recommended way was "by the EOL detection and ballast shut-down".


By the way in the same chapter was requirement, then such protection should not (on multilamp ballasts) shut down with the bad lamp more then one good one. I think this lead to only 2-lamp ballasts being offered, as the protection implemented in it's reliable, but simple form always shut down the inverter - that mean, then for more lamps the ballast box shall in fact have multiple independent ballasts, what make their integration into one box not cost effective anymore. May be the electrode fusing based protection with parallel lamp branches (each having max 2 lamps in series) connected to the one common inverter might fulfill this requirement, but the reliability of the electrode fusing method preventing "socket overheat" is in question there.

I can understand why they'd do EOL protection (have even seen it on some cheap 'shoplight' fixtures), as Medved mentions there is a safety issue ... Although i have never had a socket melt, i've had a couple times where its gotten so hot it actually melted a little spot in the outer glass (which the lamp end then cracked after cooling off...luckily in those cases the lamp didn't shatter or implode, (as these were lights just running inside)- just a small crack)

I have the problem of trying to source an electronic ballast to try this with, all the ones in Europe have EOL protection the only exception being CFLs so if I want to run a lamp to the point of vacuum loss I have to gut out an old cfl and hook the ballast up to the dead tube. This kills the ballast as well as the tube so I only get one shot at it.

Most CFLs have end of life protection too now though so it's getting increasingly difficult.

@TiCoune66, I use a ballast designed for a single F48T12 lamp as my tester/trasher ballast.  I have it mounted in a F40 strip light fixture.  It will run em in the ground!!!  Here is my video of one run till it loses vacuum:
http://www.youtube.com/watch?v=ymWE0KQRVr4

I have a slimline ballast that is "half-dead", but seems to run lamps in parallel. Also I think the good side for vacuum side is the working one. I could try but is it going to overheat the ballast, even with small lamps?

...the only exception being CFLs...

Most CFL's i've seen used at least the "filament fusing" protection concept, but the ballast is not designed to reliably survive this, so very frequently the ballast die just before the filament break. The protection is here meant only to avoid excessive heat buildup, but not to keep the ballast working.
An exception are "unlimitted ON-OFF cycles" types (e.g. Osram Facility), what usually use quite sophisticated ballast (per CFL "standard"; it usually contain the same controller IC as is used in "mature" high quality fluorescent ballasts, so include preheat timer, that always correctly preheat cathodes even after very short OFF-ON cycle and reliable EOL protection).
But even these suffer from the fact, then ballast components are designed for similar life expectancy as the lamp, what mean there is not much lifetime remaining on the ballast (mainly the electrolytic tank and lamp resonance capacitor and even coil's bobbin material get quite crispy due to the long heat exposure)