And you thought ESD was bad...
Try mains arc flash into your micro!
Arc flash is crazy. It's not so big a deal around these parts -- USA 120V just kind of goes pop, thump, dead. 240V+ is where things get interesting. It's not "interesting" enough to jump out at you (that's where "medium voltage" begins, technically 600V+), but once it's jumped, it does like to keep on going!
Some years ago, I'd designed an industrial 10kW inverter module. 480VAC input, so about 650VDC total supply. This used SOT-227 "minibloc" modules, four in an H-bridge, screwed to the bottom side of the PCB. I was testing with IGBTs, to see if they would be suitable at lower switching frequencies (the module was designed for MOSFETs). I of course discovered they weren't suitable, and the aftermath was interesting to behold.
What happens in a situation like this, is:
1. Starting from the time a fault happens (in this case, a device failing shorted from excessive heat dissipation), in about 20 microseconds, it's past the point of no return. It can't turn off, no matter how hard you try. It becomes an ever-lower value resistor.
2. After about 100 microseconds, the die and bondwire melt, then vaporize. (Effective resistance rises, because it's now an arc.) Gas production begins. Soon, the case fractures.
3. After a millisecond, the case explodes, and the gas expands. The gas is hot with assorted metals and organics from the connections and packaging material, and rich with ionization from the arc. It's very hot, and very conductive.
4. The expanding gas touches the PCB underside. The PCB is cold, so vapors begin to condense on it, leaving a powdery black (and also maybe gummy, from organics) residue.
5. Where the PCB is conductive (uncoated areas: pins, pads, untented vias), and any of those sites have an opposite polarity from the expanding arc, the arc glomps onto them as well. We now have inadvertent arc electrodes...
6. The arc continues to draw fault current from the supply (capacitors are discharged by now, but mains is still flowing). This superheats the air around the arc, vaporizing more material, and melting the "electrodes".
7. After 1-100ms, the fuse finally blows, clearing the fault. (In this case, it would've been a couple ms, because semiconductor fuses were used. In the worst case, the mains wiring breaker might open in 10-100ms.) The arc's ionization dissipates, the gas cools, smoke condenses onto nearby surfaces, and whatever remains in the air, bellows out from the area with that characteristic smell.
Tearing it apart, I observed a nicely cratered SOT-227, plenty of black smudge, and erosion and cratering of nearby traces on the PCB. One copper pour was actually burned away several mm, probably beginning at a soldermask pinhole. A blobby melted copper edge was left. There was erosion around a capacitor pin, as well, though it didn't melt (it was too thick for that, I guess).
All in all: expect everything in the vicinity of an arc flash to become conductive. Expect any exposed metal to become a target of violence. Arc flash indiscriminately fucks shit up. And, yes, needless to say, that includes the technician, so don't forget your PPE, and guards and all those precautions, when working on industrial equipment!
Domestic appliances aren't quite so violent, but this microwave oven case demonstrates how things can still turn nasty, on smaller scales.
Tim