Ok, things are not as bad as they seemed at first...
1) The -ve going spike must exceed the battery voltage (16V) before the diode conducts. That seems accurate. Then there is a 8V spike over the batt V with a bit of decayed ringing (while the current [blue trace] continues to collapse] which must include the Vf of the diode. Perhaps we're looking 7V of 'extra' voltage, some of which will accrue to parasitic inductance and lead resistances. Guessing at perhaps 2 or 3V due to that ( the pulse current flow can be 250A + thru the diode based on a 0.1Ω current sense test I did earlier). That leaves 4V unexplained. I am wondering if Vf change at high pulse currents can account for that. I see no spec on that in the spec sheet.
The spike induced failure event.
The unsnubbed >50V spike was progressively killing an avalanche rated NFET (60 Vds, 4Vth at ±20Vgs spec.) that handles battery discharge loading for battery internal resistance tests. I have since added an 18V Zener Vgs clamp to that NFET. I am guessing it's possible with all the fast edge, ringing high currents circulating that parasitic inductance & capacitance in the assembly could cause some Vgs hammering. I noted that the (40Vce) NPN 2n3904 (efollower push pull) in the NFET gate drive (10Ω gate resistor) started to 'leak' 1.7V thru its collector- emitter after the NFET failure (Vds short). That implies the gate voltage spiked (over 40V) somehow.
Also interesting is another identical NFET used as a drain - source battery polarity protector in the same discharge circuit was fine. Its gate drive is fed by a 18V amplified zener regulator as it is directly exposed to the kickback pulses otherwise.
UPDATE:
The installed snubber appears to be doing the job so far. Max bounce (parasitic drain gate capacitance) on the FET gate is a bit under its Vth. The actual snubbed kickback excursion at the FET terminals is 24V (8V spike + BattV) when all the parasitics are in play.
Comment:
The snubber design is a bit unusual in that it had to accommodate a possible reverse polarity connection on the battery. Thus a stock freewheel diode could be destroyed in that event. The diode is now in series with a low ESR 1000uF,50V DC blocking cap. In parallel with the cap is a 0.4Ω , 2A PTC. The PTC discharges the cap during operation and acts as an E-fuse in the event of batt. polarity reversal, thereby protecting the pulse diode. It works, I tried reversing the batt!
Compared to using a bank of 4 or 5, bipolar 1.5KE series TVS units with 1 ohm limiting resistors, this solution snubs a lot better and makes a lot less heat!