Thinking some of it is staged, but the idea is plausible.
The mains network is made of wires, moderately well spaced, which therefore exhibit transmission line characteristics. If anything like good old U.S. "Romex", it'll be on the order of 100-150 ohms.
A spark is extremely fast, mere nanoseconds on the rising edge. If produced from a capacitor discharge, the peak current can be ridiculous: hundreds of amperes!
At the moment of ignition, the spark delivers a current equal to (HV - Vspark) / Zo. For a 50kV spark, this could be over 300A.
The wave launches down the wires, bouncing off stubs (such as cords plugged in), and eventually dissipating in the losses of the cable (which is usually PVC insulation, quite lossy at RF) and whatever loads (transformers and light bulbs should be reasonably lossy, if imperfectly matched).
The wire likely has a pretty high breakdown voltage. Sparks should occur near the first break or opening in the wire, usually where something is attached.
Light fixtures (using something like an Edison, pin or bayonet base) have the terminals relatively close together, so would be prime targets for sparking. Junction boxes may be as well, if the insulation is pulled tight around a sharp corner, or the wire nuts or screw terminals are rather close together, or near ground.
Appliances will vary. Most equipment with an EMI filter will shrug off this kind of abuse: there's a big film cap facing the outside world, which will reflect the spark as a short circuit. Filters which don't have a capacitor in this position, will probably see some sparking around the common mode filter choke.
Sparks to ground may be more dangerous than sparks between the lines. EMI filters can't afford much capacitance to ground, and often have the CMC facing the line, without Y-caps from line-input to GND. Thus, a spark will jump across the CMC, since, that's its purpose, it's suppose to allow a large CM voltage drop!
As for energy delivered, I doubt it would be much in a 100/120V environment, but with 240V over there, there is a greater capacity for arc flash, or something like it. Absolutely, the ionized (spark) path will sink some mains current, especially if it was ignited during a mains voltage peak.
Two industrial applications come to mind:
1. Welding and plasma cutting, where a high voltage starts/assists the low voltage, high current, arc (the HV is usually pulsed HF, generated from a Tesla coil-like circuit).
2. EDM (electro-discharge machining), where the cutting electrode is powered by an aggressive foldback power supply. The open-circuit voltage is perhaps a few hundred volts. When a spark is initiated (in this case, on its own, by bringing the cutting tool into near-contact with the work), the voltage drops rapidly, and the current is increased substantially. Over a longer time scale (~us to ~ms?), tens or hundreds of amperes are dumped into the cut, obliterating the contact area. (A nimble actuator also withdraws the tool slightly, stretching the spark and preventing continuous arcing, or welding.) Over time, eventually the whole work surface is formed into the profile of the cutting tool (usually a tungsten or graphite face profile, or a very fine copper wire for making better-than-laser-fine cuts).
Tim