Yes, they can! (I don't mean burn-in period during which some fluorescent tubes are dim in the centre.)

I've got a 3-LED Panlux lamp for about 3 years. It didn't light up occasionally (discussed here) but the problems have ceased. Still in the same socket at the same place but has worked flawlessly for a year - turned on several times every day. Actually the problems have ceased after I posted about them.
A similar experience with a totally dead CFL lamp - came back to live when tested the last time before carrying it to the EOL CFL box in our supermarket. Worked a few more months after that!

I imagine an electronic part may have borderline parameters that can change in either direction. If there are more of them, it could make the system stop working and then start working again.

An off-topic example could also be our washing machine. During about first two years it occasionally stopped washing for 20-30 minutes with LEDs flashing in the way not covered in the troubleshooting guide. Surprisingly, when its 2-year warranty has ended, these problems have also ceased. Now it's almost EOL after 10 years of use but because of corrosion and not due to electronics!

I wonder if anybody else experienced a lamp that fixed itself and came back to life after being untouched for some time?

you can repair filament in incandescent lamp by rotating it.
For electronic, this may be caused by weak soldering.
For cfl, internal filament interrupted was joined again by an shake.

Not lamps, but generally electronic. I would name it Murphy's laws at work: Way too often something starts to behave erratically, so I have prepared the tools to fix it. But just after that the thing starts to work correctly again, so it has convinced me to not disassemble it this time yet. Well, my roommate at college used to say, such things just need to be threaten by the screwdriver in order to work again. He used an "explanation" for that: "It sees you are determined to really fix it, so it concludes that way of annoying you won't work, so saves it's effort for sometime in the future".

I have seem some...just can't think of any.  I know that a shop light in our basement that is hooked inot the main lights either works or doesn't work sometimes, but whenever it does wook it works perfectly, and when it doesn't, you can't even tell it has power.

Also seen some electronics do that...but I can't think of any again...

And while it wasn't exactly untouched, I managed to put this back together: http://www.lighting-gallery.net/gallery/displayimage.php?album=2545&pos=0&pid=96920

I had one case that i fixed, but if i would not, it _could_ have fixed itself at some point

When i lived in a dorm room for some period, one day the ceiling light just decided that it won't start anymore. Flip the switch, Nothing happens

Ceiling light was a 36W T8 fluorescent, Switch Start, lantern was less than a year old

After all the basic tests (presence of voltage, swap lamp and starter with a working lantern, ...) i found out the ballast is open circuit

What happened :

In the ballasts based on the Vossloh Schwabe design, as the Eltam L36.105 in that lantern is, during manufacture, they crimp copper ends on the wires coming from the coil. then dip the entire unit in polyester, then install a terminal block on the copper ends

It is assumed that during installation of the terminal block, it will break through the layer of polyester on the copper parts and make contact. Well, i happened to get the one of the lot in which this did not happen

Apparently the ballast worked for a while with the contact being made in a tiny spot, then (after so much heating and cooling cycles) lost contact

I solved the problem by taking off and putting back on the terminal. But if i would not, i guess it could have eventually arced at the point where contact existed before and re-welded itself, maybe for a while and maybe forever

I once had an under cabinet fixture with a hard wired glow bottle starter. The starter, after not much use, got stuck. I put the light away for a few years and when I went to use it again, it somehow got unstuck. I did not attempt to repair it neither when I put it away nor when I got it out of storage.

I guess the starter have recovered once it have cooled down really completely (or to lower temperatures than before). All that meant larger forces to pull apart the microweld formed in the contact spot...
I guess this is one of the very few situations, where it could be explained in really rational way and not just as a coincidence...

Yesterday, this lamp suddenly turned off after being lit for about an hour. I haven't noticed any power interruption, in fact even the UPS feeding two computers in the same room hasn't clicked.
First, I tried my table lamp and it worked. Then I went to the (mechanical) light switch, turned it off and turned it on again. The lamp lit up and worked stably for a few next hours.

Probably the electronic ballast is near its EOL after about 20,000 hours. I wonder if the incident reoccurs or the ballast has fixed itself.

Yesterday, this lamp suddenly turned off after being lit for about an hour. I haven't noticed any power interruption, in fact even the UPS feeding two computers in the same room hasn't clicked.
First, I tried my table lamp and it worked. Then I went to the (mechanical) light switch, turned it off and turned it on again. The lamp lit up and worked stably for a few next hours.

Probably the electronic ballast is near its EOL after about 20,000 hours. I wonder if the incident reoccurs or the ballast has fixed itself.

That looks like either the switch is faulty, or the lamp is at or near the EOL and triggers the EOL protection (Opple uses a thyristor shorting the base of the upper transistor to the negative rail, so shutting down the ballast and preventing it's restart, when detecting higher voltage on the lamp for longer time. It is reset by disconnecting he power).

I have a 15 watt pygmy bulb nightlight in my bedroom some times when it stops working if I pick it up it will often re weld the filament end and the lamp will come on oven brighter than before it will usually last all night like that so I got time to go buy a new one. Incidently occasionally when a bulb blows it will take out a 3 amp plug fuse

@Medved: I don't believe in the faulty switch. It's a classical wall rocker switch design used for decades. It works or it's broken but it wouldn't cause that sporadic failure. To be sure, I checked the terminals and they were perfectly tight.
The lamp near EOL could be a bit more realistic scenario. IIRC, it's the third (or fourth?) lamp in that fitting, so the lamp life has always been at least 3 years. This lamp has been installed in the last spring, so it should be still fine. But who knows...

As for shutting down the ballast - I can clearly remember that the last lamp EOL was accompanied by a few turning OFFs and ONs (within about 10 minutes) until the lamp finally passed away.
Couldn't the ballast be turned off by a peak in mains? The APC UPS hasn't intervened probably because of my user settings

SENSE    : Low
LOTRANS  : 160.0 Volts
HITRANS  : 270.0 Volts

adjusted this way because of too many unnecessary interventions with default settings. Yeah a slight overvoltage might be present, current reading is

LINEV    : 244.0 Volts

The good news is that the sudden blackout hasn't reappeared yet.


@hannahs lights: Yeah, low wattage filaments can be fixed quite easily in the way you described. Not so easy with higher wattage ones since the coil shrinks and you can't do the re-welding.

I don't believe in the faulty switch. It's a classical wall rocker switch design used for decades. It works or it's broken but it wouldn't cause that sporadic failure. To be sure, I checked the terminals and they were perfectly tight.

I guess you mean the type made from somewhere 50's till today, before 89' the only one you can buy.
I've seen many times, before they fail for good, they sometimes make the bad contact for a while (it is, when the end with the contact is just about to snap off, it bends there, so looses contact pressure).

The lamp near EOL could be a bit more realistic scenario. IIRC, it's the third (or fourth?) lamp in that fitting, so the lamp life has always been at least 3 years. This lamp has been installed in the last spring, so it should be still fine. But who knows...
As for shutting down the ballast - I can clearly remember that the last lamp EOL was accompanied by a few turning OFFs and ONs (within about 10 minutes) until the lamp finally passed away.
Couldn't the ballast be turned off by a peak in mains? The APC UPS hasn't intervened probably because of my user settings

The ballast needs few seconds power OFF for the capacitor to discharge so the thyristor switches OFF (it is the "<100uA holding" type, biased by ~100kOhm to the positve rail, so with 10uF capacitor that equals to more than 3.5sec), only then you can restart it. Reducing the voltage to 100V won't switch OFF the thyristor, you really need the power disconnected (<10V). So I doubt the UPS won't take over with such long interruptions.

But it could be a bad solder joint on the main electrolytic: The ballast will operate somehow, but the protection will become more sensitive and the lamp age faster. But once the protection triggers, the missing capacitor in the circuit may mean it sometimes reset on it's own, so allows the ballast to restart.