Also, why is it more of a problem on 3-phase services compared to single phase?
It's in the timing of the peaks (both positive and negative) of the various phases. In a 120/240 single phase system the positive peak of one side happens at the same point in time (with respect to a cycle of 60Hz) as the negative peak of the other side. This, assuming even loading of both sides, the neutral will not see much if any current. Computer switching power supplies only draw current at the voltage peaks (both positive and negative) of the phase they're connect to. In a 3 phase system the peaks happen at differing times over a cycle of 60Hz, so the current spikes on a neutral of a Y system will not cancel out. (overlap of the current spikes is unlikely). The neutral will see (current spikes on phase A)+(current spikes on phase B)+(current spikes on phase C). If the loads are evenly distributed the neutral will see 3x the current. New rules in Europe will be requiring these power supplies to draw current more like resistive loads (low harmonics, good power factor) so this overloaded neutral issue should ease.
