I am curious about this. With these capacitors, is it the case that power cycling them does them no harm? I can see that they should be kept cool, or within their optimal operating range either by design or additional active cooling. But does power, and therefore heat cycling really have no effect on their long term life span? Seems odd to me.
Why - like all things for something to reduce life there has to be a wear out mechanism.
So for example, light bulbs and tube filaments. At switch on, they are cold and hence low resistance resulting in a high current until they get hot. This causes a wear mechanism via actual thermal expansion/contraction. But also the filament gradually gets weaker though evaporation and vibration.
With semiconductors, the devices can be damaged by exceeding a maximum temperature on the diode junction (Tjmax). Most rectifier devices I've seen have a repetitive pulse current rating (Magnitude and repeat rate on a curve) which if kept below will not result in a damaging temperature.
Electrolytic caps of the type used in power supplies are large and dont need to have very fine connections that might get hot during inrush - though I guess there may be thermal stress on them. However by far the dominant form of lifetime limit is from the (wet) electrolyte drying out. This is accelerated by heat. If they are operating they will be internally dissipating heat (in the ESR) so even if you add fan cooling they will be hotter than when switched off. So you are "consuming" their life all the time they are on. As I pointed out above - lifetime typically halved for each 10c increase in temperature.
In a former life I developed variable speed AC drives of up to 95kW. These have huge capacitor banks driven by a rectifier. Due to the size of the capacitor bank it was necessary to have inrush limiting which was a resistor shorted out by a relay after the capacitors were charged (typically only a few seconds). One of our design tests was an accelerated life On/Off cycle test. The Test unit was switched on and off as often as possible with the capacitors dischared each time. This was typically on a 30 second cycle. The test protocol required 100,000 cycles - it took 1-2 months to complete.
In the 20 odd years I was involved we never had a failure from a capacitor. If we had a test fail, most commonly it would be the inrush resistor due to being under-rated, very rarely we had the rectifier fail.