The other thought I had (and my wife too!?!) we saw a program about a luxury building that had to have special equipment to combat the possible fire problems associated with lots of dimmer controls around the building.
Ah that sounds like Hamster's program on the Burj Al Arab hotel (err, sorry, Richard Hammond's Engineering Connections) - nice program but I thought his explanation of that issue was absolutely rubbish.
To explain it briefly, our power distribution systems are three phase down to street level - and into the building once you get to anything but the smaller commercial ones. Each phase is 240V to neutral, and for small loads that's all you use. The beauty is that by spacing the three phases 120˚ apart, a balanced load will sum to zero in the neutral. When one phase is at it's positive peak (ie +1), the other two are at -0.5, when one is at it's negative peak (-1), the other two are at +0.5), when one is zero, the other two are at + and - 0.8<something>. The sum of three sine waves, 120˚ apart, and of equal amplitude is always zero.
What this means for power distribution is that the neutral current is NEVER higher than the current in the most heavily loaded phase, and is almost always significantly less. By the time you average enough loads spread over multiple phases, as is the case when you get back up the distribution system a bit, is that the neutral current is always going to be a fraction of the phase currents and so you can make the cable considerably smaller - the neutral is the single cable along the top of the pylons.
That's the case for resistive loads, and also for loads with a purely inductive (motors) or capacitive element. Once your phase current ceases to be a pure sine wave, then this whole assumption breaks down. It came to the fore several decades ago when large office blocks had lots of flourescent lights and started getting IT equipment. Both have a poor current waveform, early IT gear especially so. Both tend to have a current waveform that rises slowly with voltage for a bit, then rises rapidly to a peak, falls rapidly, and then falls slowly as the voltage falls to zero - the rounded peak of the sine wave is "squashed in" at the sides and has a narrower, taller peak. As demonstrated in the program, phase angle dimmers have a particularly bad waveform.
The technical bit is that such a waveform has a large "third harmonic" component, the non-technical answer is that this component of the current does NOT subtract in a three phase system, it adds up. If all the current was third harmonic, the neutral current could be up to three times the size of each phase current. I don't think it takes too much imagination to see what this does to a cable designed for maximum of one times the phase current
In reality, the filter banks shown in the program were probably nothing to do with the internal wiring - you just size the neutrals to suit. However the electricity supply companies have rules on how "dirty" your load can be - and your supply as a generator*. Not only do they have to consider their own neutrals, but they also have to consider the effect of such "dirt" on their other customers. lastly, some of the waveforms you can get are not kind to transformers designed for nice smooth 50Hz - feed them all these higher frequency harmonics and it can cause additional losses, which manifests as additional heat (there are other considerations, but I'll leave them aside).
* I have it on good authority that the new windfarm off Walney (Barrow in Furness) could not be connected to the grid recently. They use static converters (a rectifier and inverter) to convert variable frequency from the turbines to 50Hz for the grid - and the output wasn't clean enough. Apparently they left enough room in the offshore substation for filters if this proved to be the case.
Hopefully this sheds some light on the subject.
