The turnon event will put a lot of pulsed thermal power
onto the switch device and the heat sinking has to have
enough close-in thermal mass to keep juntion temperature
peak, sane. Far-away thermal mass won't be fast enough
to save the junction at the top of the stack.
But too, turnon has to be leisurely enough that the switch is
in control of the current (despite that it also has to take the
abuse meanwhile). Experiment with gate slew control and
load dV/dt feedback to force a ramp that leads to a tolerable
max current during charging. Take a stab at modeling the
thermal stack, with masses and time constants, realistically
all the way to the D-B, drain-neck junction so that you can
estimate a safe pulsed operating area and "make it so".