Thank you very much for your analysis.
You could have an inductive source (2a) - try a capacitor to ground at the front end of your switch?
Source is usually two lead-acid 12V batteries in series with their charger. It's a quality charger but I've experienced the same failure in systems powered by a mean well 27V regulated power supply.
You could have an inductive load (2a) - back EMF? less likely but the body diodes of the FETS won't save you in this config, perhaps a diode across the whole switch?
The load is a TAS5622 class D audio amplifier. Definitely not inductive.
You could have a heavily capacitive load (1) - less likely to be a problem as you have a reasonably slow turn-on with that 50K pull-down. Have you calculate turn-on time?
That is the case. The load has about 3000uF of decoupling.
I've gone as far as adding a 100nF capacitor in series with R35A for a super-slow turn on and it doesn't seem to fix the issue.
In any case, those PMOS are good up to 200A peak current, I don't think I'm getting even closer with a 3000uF load.
You could be taking to long to switch on (3) - perhaps your turn-on is too slow leading to too long in the linear region and overheating? Reduce 50K, or try a different driver topology.
As I said, I tried way way way slower turn on and it doesn't fix the issue. Plus I would expect some physical damage in the parts if they have been killed by overheating.