So my theory justifying the above is as follows. If the relay fails, then the 400 V bus charges as normal, and until the IGBTs are switching, there is essentially no load on the 400 V bus. So there is simply no way for the charrger microcontroller to know that the input relay has failed at this point. So it gets everything ready, perhaps does some handshaking with other vehicle computers, and if all goes well it starts ramping up the charge current in a few seconds. When the charge load is less than say 2 A, then the voltage drop across the two 4.7 Ω pre-charge resistors is less than 20 V, which is less than 10% of the mains. At two amps, the power dissipated by each resistor is I²R = 2×2×4.7 ~= 19 W, a severe overload, but they can probably take that sort of power for a few seconds. Note that with 10% less AC voltage available for charging, the PFC (Power Factor Correction) stage has to work 10% harder to maintain 400 V, so really it's 2.2 A at this point.
But when the charge power ramps up to say 920 W. so the current from 230 V would normally be 4 A, then the voltage drop across the pre-charge resistors is at least 40 V. That means the PFC stage has to work at least 17% harder, so it's really 4×1.17 = 4.7 A, which really means 44 V drop, so the current is really about 5 A, and each resistor is dissipating some 5×5×4.7 = 118 W. Surely at this point, they would fail open circuit. But there is a chance that they could fail high resistance, evaporating away some of the wire (assuming wire-wound types, as they appear to be), leaving some wire intact. This means that they would get even more heat with the same current, but the AC input to the PFC stage would drop even more quickly as the charge current ramps up. When the PFC stage input voltage drops below about 85 VAC, the charrger microcontroller would detect an error and stop charging. This protects the pre-charge resistors from further deterioration, so they might stabilise at a certain much higher than nominal resistance.
So my theory predicts two different outcomes, though one is more likely than the other. The more likely scenario is that the pre-charge resistors fail open circuit suddenly. When that happens, the 400 V bus collapses very quickly, which might cause stored energy in the output inductor(s) to cause the small doghouse capacitors to explode. These often fail short circuit, hence the fuse in the motor controller blows. We've seen plenty examples of this.
The other, less likely outcome is that the pre-charge resistors fail high resistance. As indicated above, this could cause the charrger microcontroller to terminate charging so quickly that the pre-charge resistors don't suffer much more degradation. This could be the situation with ChristopheFR's Zero. Alternatively, every time you fire up the charrger, it might be like playing Russian Roulette with the charrger; eventually it dies (because the pre-charge resistors fail open circuit), but it might get lucky and survive a few starts. Maybe ChristopheFR was lucky, and/or didn't try starting the charrger many times. Until the pre-charge resistors fail under load, you don't get the sudden stopping of charging, so you don't see the doghouse capacitors explode, and the DC-DC still charges the auxiliary battery.
One thing that bothers me with this theory is that any
interruption of AC while charging should cause the doghouse capacitors to explode. Otherwise, when the input resistors fail open circuit, the charrger should quietly refuse to start from them on, and the DC-DC should continue to charge the auxiliary battery. I imagine that most charges would be via J1772, so if you interrupt the charge by taking out the connector, there is a digital signal to the charrger microcontroller, which allows the microcontroller to ramp down the current in a tenth of a second or so; that's all good. But surely sometimes the AC fails for other reasons: a breaker trips, the charge is via a wall plug and is turned off before the charge is complete, blackout, brown out, or a rat chews the power cord. I don't hear reports of on-board charrgers failing for these reasons.
The other thing is that there is considerable energy in the 220 μF capacitors, so that interrupting the AC input (which happens all the time 100 or 120 times per second due to the nature of alternating current), and these are presumably sized such that over the 8-10 ms between peaks of the mains, these capacitors can run the charger with only a minor dip in 400 V rail voltage. I realise that between peaks, the capacitors are still being charged via the PFC boost stage, but there is less current charging the 400 V bus capacitors between peaks. The current is sinusoidal, as well as the voltage; that's why it has good Power Factor. Kiev's theory about poor auxiliary battery voltage may explain this; if the IGBTs get too low a voltage on their gates, they could fail to turn on properly or at all. There might be some sort of desaturation protection, which causes the IGBTs to suddenly stop switching. However, in my very limited experience, desaturation protection is more of a gradual cut-back of gate pulse width than a sudden shut-off.
So I suspect that while the relay is the root cause of some
failures, something else is causing the doghouse capacitors to explode. In Skyogger#1's case, we could blame the poorly manufactured 39 kΩ resistor, but this doesn't seem to be a common issue, at least so far.
Let the speculation continue