Unlike fuses, varistors are wired parallel to the circuit they are protecting (from voltage spikes), so their failure does not interrupt current to the circuit. They work by shunting voltage higher than their rated set point to circuit ground (in this case the neutral mains leg), thus preventing it from reaching the circuit - used to protect from spikes due to lightning or power company anomalies.
Fireworks is occasionally the ultimate failure-mode of those devices. They are made mainly of zinc oxide, crystallized and “doped” with tin and bismuth oxides in a manner that gives it semiconductor properties such that it presents high resistance to voltages below its cutoff point and thus in normal operation doesn’t conduct. (The specific composition of doping agents and crystalline structure is fine-tuned in order to achieve the desired voltage rating.). As the input voltage rises above the device’s cutoff point it rapidly becomes conductive, shunting the excessive voltage across itself.
Although they become “conductive”, there is still some residual resistance so some of the energy of the voltage spike is converted to heat rather than shunted, and as a result the crystal structure is altered/damaged a little. They can tolerate a number of these “events” and still retain their efficacy, but each time they do so their semiconductor properties are compromised a bit, and at some point they can begin to conduct at lower voltages. When this begins to happen the device runs the risk of going into thermal runaway, and at that point its fate is sealed; usually they just quietly overheat and slowly burn themselves out with minimal fanfare (you might notice a little bulge or area of discoloration/scorching on the package if you look closely), of course leaving you with reduced circuit protection but no drama, but sometimes they do go out with a “bang” like your pal experienced (which of course can start fires and damage other stuff).
I think they are an often neglected part of restorations and ideally should probably be replaced on a regular basis. I once read an article by an engineer who suggested that MOVs should be replaced after any known spike “event” (like a close lightning strike that damages any nearby equipment) as well as when doing any service to the power supply (such as recapping or replacing diodes); they are dirt cheap and easy to replace, so it’s “an ounce of prevention” thing.