We already waste a great deal of food. Having it too cold will damage food, as will too hot. Older fridges used a sawtooth cooling pattern (on/off) but newer ones, with fancier motors, have might tighter regulation. I'm sceptical of how much energy can be scavenged.

I'm sure with enough IoT you could delay motor start +/-30s across a region to smooth the tip of a load spike, but it sounds very complicated. How would end users be compensated for such a tiny shift in load? Isn't instaneous power demand a martingale process?

It sounds like it would be easier and cheaper to have a few batteries.

I wasn't suggesting making it so cold as to damage food. I also wasn't arguing against batteries.

If you were suggesting a refrigerator to run the cooling in the night so much so that it'd be able to sustain a safe temperature for many hours while conserving energy in the day, then yes, you were suggesting making it so cold as to damage food - a few degrees difference isn't sufficient to achieve what you want, and if your temperature fluctuates more than 3-4 degrees then the food either gets unsafely warm or gets frozen/damaged.

I wasn't suggesting that, no.

I was just generally arguing against the parent proposition as well.

Well, if you're a skeptic as to how much can be saved, you might want to look at the many "smart grid" studies, because this has been a hot topic for a while and is already implemented to a small degree.