A very simple explanation of 'Feed Impedance' is it is a measure of supply 'solidity'.
The long explanation is the feed impedance determines the extent by which the voltage will fluctuate under a varying load. The lower the feed impedance the less the voltage will fluctuate when the load changes.
However, there is a negative aspect to feed impedance. If too high (even if the variations in voltage are tolerated) then there may not be sufficient current available to activate any protection. Such a scenario happened with a radio repeater on top a hill being powered from a factory below. The power run was made out of 2.5mm² and was long enough to have a 40V drop with only 100W being drawn.
The drain above was approximately 0.5A meaning the run had an impedance of 80W. The repeater indicated it was running on batteries and the technician called to see if there was a mains failure, only to be told no and that the breaker had also not tripped. Investigation (after the many phone calls and the repeater now dead) revealed a short on a badly installed plug point at the repeater - but with 80W feed the maximum current draw was 3 amps - not enough to trip even the 5A breaker. Ok, so this was a little extreme but it does illustrate the point.
Yet the opposite is also true. When the feed impedance is very low the currents developed during a fault (i.e. short circuit) are exceedingly high. This current can be high enough to destroy protection devices (circuit breakers). This usually occurs when the breaker opens but can also happen during a fault without the protection operating. This must not be seen as an argument against low fed impedance, but rather as a case of ensuring the protection can accommodate the supply impedance.
Let us recap; Flicker is the result of a change in brilliance of a lamp as a result of a change in voltage. A change in voltage is the result of a change in load on a point owing to the impedance at that point. The lower the impedance, the less the change in voltage, the less the flicker.
This means, in simple terms, that one cannot put a "flicker value" to a non-constant load (oh yes, I've seen some who have tried!). Feed impedance is a site related figure, so having the same non-constant load at two different sites can produce two different flicker results purely owing to the different feed impedances at each site.
The argument "but the arc welder doesn't cause trouble at the office" when a neighbour complains of flickering lights, holds no ground as it is not just the amount of current the arc welder requires, but the feed impedance difference between the office and home. The office supply is likely to be a lot 'beefier' hence less fluctuation (therefore less flicker) with the arc welder current meandering all over the place.
This same argument holds true for huge industrial applications. In fact, it is probably more true than our above mentioned arc welder example. Companies cannot "just move shop" without taking into consideration the feed impedances of the new site, especially if their energy demands are as stable as a bowl of jelly during an earthquake! The bigger the ratio in immediate changes, the greater the concern over causing flicker at their neighbours.