Consider the situation when your 350 V supply is set to max output, no load. The unregulated side might be at 360 V but let's call it 400 for simplicity. If you apply a short-circuit to the output then your pass element acts as a voltage follower backed by the low impedance of whatever capacitance you have upstream, your peak current is essentially limited by Rdson because the source has been pulled down to zero and the gate hasn't moved yet. Your pass element now sees maybe 400 A and 400 V.
Hold on a second.
1. Rds(on) is a convenience, and it only applies at low voltages. Check Fig. 3.
Fig.1 is more pertinent, though it doesn't show the full range.
2. As it happens, SuperJunction MOSFETs have current saturation different from traditional MOS. Transconductance crashes beyond about 7V, so that even under a pulsed condition, more than about 40A is unlikely from this device.
IGBTs have a similar behavior, though the available transconductance tapers off less sharply, so that short circuit current tends to be limited more by Vgs(on).
I've measured STP19NM50N (from ST's SuperJunction family) which behaves this way. See for example this measured curve:
https://www.seventransistorlabs.com/Images/STP19NM50N%20Drain%20Output%20Curves.pngI didn't plot 8V because it was almost identical to 7V.
Still, this is beyond the SOA, even for pulsed conditions.
So...
Essentially this always puts you outside the SOA, and all you can do is make the current limit cut in fast enough that nothing blows up. Realistically it's probably easier to design the supply as a current source controlled for voltage.
You're just going to throw your hands up? When the solution is a resistor and transistor away? Or a resistor and a zener?
The throwaway VCCS idea is actually quite an important concept, though much more complicated than needed here. You need to go that route for a switching regulator, since the inductor's state variable is current.
Tim