SCR might stay latched on, probably a bad choice.
An "avalanche rated" MOSFET might solve the problem
cleanly, depending on the stored energy. Should be
app notes / selection info for any such parts.
A "catch diode" that returns overshoot energy to the
supply could also do the job, less dissipatively than
the TVS (TVS will have a Joule energy per-pulse
limit, and a continuous-pulse-train power limit, along
with steady state power limit, and since it drops the
full voltage @ full current for a period, this is worse
in terms of applied power than a catch-diode.
I find it unlikely that shorting drain to supply causes
an overvoltage failure mode. Overcurrent or over-
power, yes. I would not be so quick to assign the
failure mode without some inspection of the die (and
this wants more / better equipment and experience
than the hobbyist is likely to have).
If it is overvoltage then the fail mode should produce
a gate leakage signature and occur on a single pulse.
Overcurrent and over-power will take multiple pulses
to blow open any source bond wires or local interconnect
(at which point full drain voltage D-G is applied and gate
rupture could occur, without the body / neck shielding
the gate).
Rooting out the true failure series-of-events and
mechanisms might help you decide but I kind of doubt
it's necessary. If everything on the page is on one
board then I suggest the catch diode approach (like
a fat 50-100V Schottky), anode to LEAD B, cathode
to LEAD A, and forget the TVS or call it a backstop.