Mosfets failure in inverter bridge

A few years ago I had to find root cause of Mosfets failing in an inverter bridge. There were 4 low power devices all in a tiny TSOP-8 package. Part of a module for powering an LCD monitor. The failures were random, but in the range of a few thousand ppm during the system burn in. Peculiar thing was that it was always one pair that got burnt out - the right arm of the bridge. Always. All waveforms in a working module looked perfectly normal. Took several weeks of investigation, a trip to Taiwan and Shanghai and a massive effort fending off some of the most bizarre theories emanating from our own FAEs. Ultimately an incredible stroke of luck on the last day helped to nail it down.

1) Problem started with the control IC which put out a long rogue pulse just before the soft start sequence. Not all ICs did that, but quite a few. Current limit was not operational in that mode.

2) There were two suppliers for the inverter transformer. The magnetic design was identical but each could, and did, use their own materials. Same core geometry and number of turns and wire size etc. but you could order the core from your own vendor. One brand of core survived the longer volt seconds caused by the rogue pulse. The other saturated heavily leading to a large current spike at turn off.

3) The PCB layout was asymmetrical. The bypass capacitor for the right arm was located far away from the package with high trace inductance in the loop. The burn in consisted of repeated start ups and shut downs at different input voltages and temperatures. Three of the devices in the bridge managed to turn off safely even with the high saturation current under the rogue pulse. Only one, located on the high side of the right arm, went into avalanche during every start up cycle at high input. And when it failed it took out the low side Mosfet as well to complete the mystery.

Not sure which button I should check here ... but many of the failures I have seen come from unexpected combinations of different issues.