How do multi-voltage (120-277) thermal photocells/controllers work? If they just use the standard CdS and thermal switch, it would burn up if used on a higher voltage.
If it uses a power supply to power the CdS sensor, you could just use fully electronic ones.

I have not heard of thermal wide voltage ones (my first thought of multi voltage photocells is about electronic ones), so im guessing here

Good CdS sensor is changing its resistance over quite wide range - In light its resistance is low and in darkness it is well enough high. Lets start from darkness : It is high enough to be allmost open circuit, good enough for wide voltage range

In light, the heater is designed so 120V heat it enough to open the contact. Nobody prevents from building the heater sturdy enough to withstand 3x the power (208V), 4x the power (240V), or 5.3x the power (277V) - It heats more, but it can be made from materials that withstand the heat. It will just bend the bimetal more far away from its original position, and it is still possible to make the mechanics to allow the bimetal to go that far without getting stuck into other components

The 4x the power looks like a lot, but it is not. The cell mechanics are not a fine precise system, except the setting for the switching light level which is done by trial with the resistor values, springs in the relay, and finally tweaked with a screw

This method would work, though it would be quite crude and waste more power at high voltages, as well as have inconsistent switching levels (lower Lux when powered by high voltage) so i assume it is better to choose a cell for the specific voltage with the thermal ones

Sounds like the light switching sensitivity would change slightly at a higher voltage.
Here's what I'm thinking of.

Thermal switches actually use two bimetal stripes acting against each other (so the ambient temperature variation is compensated), but only one of them is exposed to the heater. Therefore the ambient temperature is compensated out.

So a plausible arrangement:
Using a second heater (4x the resistance compare to the "sensor" leg heater) on the "compensating" leg, connected to the full supply, causes the voltage to be compensated as well - when the CdS resistance is the same as the main heater (so the desired threshold), the sensing heater has power dissipation of (Vmains/2)^2 / Rsense.
The compensating heater dissipates Vmains^2/Rcomp. With Rcomp = 4xRsense it means the same power dissipation.
When the two bimetals are equal, the same power dissipation means the same bending, so still staying on the same threshold even when the voltage varies. So when the Rcds>Rsense, the Sense leg will be always colder than the compensating one, so keeps the light ON regardless of the mains voltage. The same when the Rcds<Rsense causes the sensing leg to be hotter, so keeps the light off, again regardless of the mains...

But not sure, if this was ever used in a real life product, it is just a possible arrangement solving the voltage variation influence. Beside the need for about 2x .. 6x higher power input (for the same sensitivity of the thermal switch assembly), it needs the CdS to be rated for 5.3x higher power dissipation rated than with an equivalent single voltage model, what means way more expensive sensing element (and that does not depend whether it is voltage compensated or not, it is just related to the wide supply range).

I wouldn't think that it would be used in real life photocontrols because thermal ones are usually marketed as being low-cost.

Well, I don't think the thermal design is cheaper than the electronic even these days (the need for large CdS cell). But for sure it was the cheapest way in the past...

But there was another reason for thermals: The time delay - prior very cheap microcontrollers, it would be too complex to implement in a reliable manner (so not sensitive to board or capacitor leakages,...) in any other way...

Thermal seems to be cheaper here.
Electronic photocells could be more expensive because they can have lots of features, such as a DC relay, silicon sensor with IR filter, etc.

Thermal seems to be cheaper here.
Electronic photocells could be more expensive because they can have lots of features, such as a DC relay, silicon sensor with IR filter, etc.

What the "DC relay" feature means for the user/application? Why should the user care at all?
It is just a marketing "blahblah" - the application does not care at all how the switching work, if it works, that is what matters. The DC relay itself is way cheaper than an AC one (way simpler - no need for shielded poles to hold the contact closed even during coil current zero cross or so, smaller parts,...).
But of course, marketing wizards are very creative in promoting any cost cutting measure as "an extra feature", even when it makes no practical meaning for the user at all, sometimes even when it actually degrades the real product value (e.g. a "brake assistant" in modern cars actually means the ABS, once activates, is not capable to track the pedal "requirement" information, so instead it then brakes at full force the actual adhesion allows; it only switches OFF once you completely release the pedal - so once the ABS activates e.g. on a small ice patch, the car then goes to full brakes, even when you wanted just a soft slow down).

The "IR filter" does have effect on the application (less sensitivity to an artificial light), but in reality it cost nothing. It is intrinsic silicon nature to be more sensitive in the IR; on the contrary, for the visible light sensors such extra IR sensitivity has to be extra suppressed (therefore the IR blocking filter in cameras), so an attempt to make it responding really on the visible only (such as CdS does) would make the sensor actually more expensive. But this at least means an improvement for the performance the user really sees (less false responses for an artificial light spill)...