Great thread everyone. I thought it might be worthwhile for some to share a real-life experience that proves this MarkT comment in a very instructive way:
Under extreme failure conditions a fuse wire vaporizes to plasma, and the copper vapour/plasma still conducts until it cools/condenses
Back when the 8080 was the norm in industrial control systems, the company I worked for was a machinery OEM using a well respected manufacturer's control system that had utilized plug-in discrete I/O modules, both analog and digital, that plugged into a local controller base, 32 I/O per controller.
The AC output modules were rated for 2 amps per channel, 85 to 240 VAC. These used IIRC an old MAC11 series triac driven with a standard optoisolator, along with usual dv/dt suppression and other things you'd expect in an industrial device, one of those things being a UL listing. The module required no external fusing - it was provided internally by a carefully designed small length of PC board trace calculated (and tested by UL) to open when the fusing current was exceeded. The fuse link side of the circuit board, which had the usual green solder mask, was protected by a piece of thick insulating paper between the board and a custom flat plate aluminum heat sink tab for the triac. Two of these boards were mounted inside an injection molded case with a plug base on the bottom, about 25mm square by 100m high. All in all, it had all the appearances of a reasonable well engineered and competently manufactured device.
We had used this system for about five years here in the USA without incident. Customers fried the AC output modules on a fairly common basis with failed solenoids and other failure mechanisms, they'd just plug in a new module and life would continue as before.
We sent several of these systems to Spain where the only differences were that the machine control voltage was 240vac. That means the currents were half of the original 120 volt devices so what's not to like? We were well within our current limits. It only took about three months to find out all was not well.
The short story is that I very quickly ended up in Madrid to figure out why everything, and I do mean everything, stopped working. Every I/O module, every processor, every power supply in that system had failed in one way or another. In all, about $30,000 worth of hardware failed all at once, all traceable back to one event.
A water control solenoid had been accidently shorted, causing the fuse link in the AC output module to blow. When it did, the higher 240vac control voltage was coupled through the plasma ball that had formed from the copper and tin of the fuse link blowing open - coupled back to the common +5 volt power supply and then from there, to every other circuit element in the system.
A complete, system wide release of the smoke because the original design engineer had saved $0.50 on a fuse, ultimately, I found out because he didn't want to change the design due to the size of the fuse. I never did get a good answer from UL why fusing wasn't required, my boss had gone cheap as well and said that if UL didn't say we needed it, he wasn't going to put it in.
The bigger problem for us was that we had no place to provide external fusing. The thought of rewiring things and adding series fuses to every output appeared impossible. The equipment manufacturer said it was our problem, they had their UL listing and UL didn't care. Tough spot to be in. The customers machine was down for several weeks while all of this went on.
Ultimately, I used 2 amp Picofuses inside the module to solve the problem. Cut out the fuse link and replace with a real fuse. Funny thing was my boss would only allow me to do the mods for the 240 volt destined modules, he gave those a different in-house part number because of the cost difference.
What I learned from that experience was priceless. Thanks for triggering the memory MarkT.
