Inverter
I don't think those old PWMs have much for anti-shoot-through
and maybe there's something going on that makes cross-
conduction happen. For example a synchronous-rectified design
that lets the negative undershoot exceed -0.7V will light
up the body diode / BJT and it could fail to turn off in time
for the next edge (a MOSFET on resistance problem, becomes
a destructive problem).
If you get trouble when the loop is closed, start by opening
the loop and try to drive it into failure while observing the
kind of things that yield clues (Vds, Id, gate terminals as
indicators of cross-conduction, etc.). There must be a steady-
state or oscillatory condition of failure and you work forward
to where you can replicate it at will, and then backward to
the root cause.