Why bother?

[Rob Strand]

It's a valid point.  When we use lux and lumens units those effects are already taken into account (at least to first order).

FWIW,  in normal operation (and at their rated voltage) incandescent bulbs all have about the same lm/W.   Higher power bulbs produce more lm and smaller bulbs lower lm but the lm/W stays the same (in reality only crudely).

(Like you say) when the voltage changes and the temperature changes the lm/w drops.  This is fairly evident in lm/w on that graph I posted.  The lm/w drops off as the applied voltage is reduced.   So a high powered bulb throttled back to 1W will produce less light output than a 1W bulb operating at it's rated voltage.

[Vivek]

So a high powered bulb throttled back to 1W will produce less light output than a 1W bulb operating at it's rated voltage.

and different color temperature too, I believe.

[pinkjimiphoton]

i gave one of these away a couple years ago. i never really spent any time inside it, but as i recall it was a very thick, chewey "chips-ish" phase shift that had a nice wobble like a univibe and the harmonics affected the intensity, or the "width" and chewyness of the effect,  reminiscent of one of them contemporary norlin maestro foot phasers, the one with the rotateable wheels on the sides of it, but then, i may be thinking of something else... so soo many circuits... so little time...

[Rob Strand]

So a high powered bulb throttled back to 1W will produce less light output than a 1W bulb operating at it's rated voltage.

and different color temperature too, I believe.

Definitely, a 1000W bulb operating at 1W would not be visible as most of the radiation is infrared.
Your eye is like a band-pass filter (the CIE curve) for the visible light frequencies.   The radiated energy
is the same as a 1W bulb but the radiation is in the low frequencies.

The lm and lux units try to measure what you see they effectively pass the radiation/light through a band-pass filter.  If the radiation frequencies don't line up with the eye band-pass filter the measurement is low.   lm and lux are numbers which match what we see.

LDRs crudely approximate the eye spectral response.    Not exactly but they have similar filtering behaviour.

The non-linearity of drive vs light level is something which comes into play in the univibes.   That's where the LED builds don't quite match.

[brett]

Hi.  I'm here to agree that it controls the clipping in a variable, smooth and almost noiseless way.
As to "How much does it control the clipping?" it helps (in my amateurish opinion) to look at the input impedance of the next stage (which 'competes' with for signal/voltage).
If the input impedance is 100k (as an example), and the bulb is fully lit and the LDR is 2k ohms, then it's unbridled clipping.  At the other dnd of the control, with a dark bulb ans 5 M ohms of resistance, there's not much clipping.  slmost none.
When the control is such that the LDR is within +/- 10 x of the input impedance of the next stage, there'll be maximum control of the clipping.
Although this looks mathematically sensible and will show up nicely on an oscilloscope, it seems that subtle changes to clipping aren't much of a change in tone.  eg turn up the gain knob on a Tubescreamer a little bit... almost no change. 
Does any of that make sense?