Mercury vapor lamps last 6-7 years. 

High pressure sodium lamps last 3-5 years. 

Probe Start metal halide lamps last 2-3 years.

 

The switching cycles do not cause any extra wear, but the high currents associated with switching ON the cold lamps create extra forces stressing the filament, so it tend to break the filament earlier. In other words, when not subjected to those high currents, the filament really break after more extensive wear, when the weak spots went even weaker.

This effect does not affect as much the thicker, higher current and/or shorter filament of low voltage lamps, but the low current high voltage lamps (what the nightlight bulbs are) are more fragile by itself, what make the effect somewhat visible.

And even then, externally reducing the inrush current does not prolong the median life as much, but it rather steepen the mortality curve and reduce the early mortality (it only make the lamps a bit more tolerant towards weak spots in the filament).

somewhat but not too noticable

There are so many variables. Any light source could potentially vary in life from a fraction of a second to over a century. The one thing - in average commercial and household use - that I have noticed, though, is that the lifespan of light sources - from incandescents, LEDs, fluorescents and HID sources - has shortened significantly within the last decade. That points to quality issues.

I've had a handful of lamps that have surprised me. An HPS lamp that lasted over 10 years dusk-to-dawn. A magnetic CFL that did over 24,000 hours. An R-30 incandescent rated for 2000 hours that lasted 10 times that (though was often dimmed). A neon NE2 on a power strip that has been burning 24/7 for over 2 decades, still bright & steady.

On the flip side, I had the Sylvania CFL that burned out in 45 minutes; the GE A-19 incandescents that burned up in 2 seconds; the NE2's that were flickering right out of the box; the HPS and mercury lamps made in China that lasted 3 months before leaking their arc tube's contents. The GE Powergrooves that lasted only a few days before losing vacuum. Yep - all defects - not a big deal on a large scale commercial installation, but a bigger deal to the residential consumer.

At our church we have incandescent sconces that burn 24/7 (code) in the main sanctuary and they can last maybe a few weeks to a few months in time ( havn't kept a very eye on them).  We had one bulb that lasted about 4 months before going, it was the last original bulb burning 24/7 that was put in there.  It did roughly 2,700 hours.  When you come in you never know if one will be out or not, one whole sconce (3 lamps each) was completely dead.

The 2 4 lamp 4ft. T8 troffers (with Phillips Altos) have been going strong for the same anount of time.  They only got turned off once because of a blown breaker.

It does very with product quality I have seen. 

moojay you prolly need some 20,000 hour aero-tech bulbs in these scones lol.

I sorta doubt it

There is that way to dim an incandescent by connecting it with diode (1n4007 etc) so that it lights in 1/2 cycle. On 240V 50 Hz the thin filament in small (40 and down) lamps suffices to cool down in the half cycle to give a very pronounced 50 Hz flicker. If each such cycle would take a significant art of the lamp's life, the lamp would perhaps not last very long... (it makes 1000 heat and cool cycles in just 20 seconds)

But this may not be very good estimation because in a lamp on diode the filament does not heat to full temp (can tell by thee color temp of such lamp which is quite low) an perhaps does not cool to full zero either in those cycles, so the cycle does not span on the entire temp range like in a lamp that is really switched on and off - maybe the worst is not in this range