Mosfet as Switch....

[FvM]

I believe that it's possible to achieve maximal fast turn-off if needed. You need to know the coil inductance and switching frequency to calculate the zener rating.

The OP didn't explicitely ask at all for faster turn-off than achieved with a regular flyback-diode. The usage of a varistor may be understood in this way, but it's just a guess.

 

[crutschow]

If using a TVS diode, do I need 1N4007 in series?
Is P6KE (6.8-400V, 600W peak, 5W steady) okay here?

What wattage of 1.5 ohm resistor would be fine?

You don't need the diode if it's a bi-directional type. Note that you should't use one with a higher voltage than a P6KE20A for a 20A surge.

For low-duty cycle a 5W resistor should work fine.

- - - Updated - - -

..................................
The OP didn't explicitely ask at all for faster turn-off than achieved with a regular flyback-diode. The usage of a varistor may be understood in this way, but it's just a guess.

True. But the OP may not have been aware of the possible degradation of the contacts from the slow turn-off that a single diode could cause.

 

[abuhafss]

As per LT Spice simulator

With single diode the surge is 31.5A and duration is 0.68s
With 1.5 Ohms resistor & diode it is 27.6A and 0.24s
With 2.2 Ohms resistor & diode it is 18.8A and 0.20s
With TVS SMCJ33A it is 31.25A and 0.45s

I doubt if 1N4007 (1A) can bear 18.8A for 0.20s?

Edit: With 22 Ohms resistor & diode it is 1.9A for 0.07sec. This seems to be more practical but, what wattage for the resistor?

 

[D.A.(Tony)Stewart]

Automotive switch requirements are much harder than you think with 2x V+ , V- and load dump surge tests, so rather than "reinvent the wheel" learn from high or low side switch solutions for automotive market, as I previously suggested.

In the case of 1N4007 , the transient will be Ip= V/R for duration T=L/R absolute max spec. The abs. Max.=I2t Rating for Fusing (t<8.3ms ) = 3.7A²sec. This was intended for 120Hz rectifier surge calculations.

I agree from datasheet, 18.8A@200ms exceeds the limit but if 200ms is a transient from Pk to 1/e*Pk, this may be the Abs.Max. & Not the recommended design limit per LSpice.

 

[abuhafss]

It worked with 2.2R/5W and 3A diode.
But, the result is not consistent......the source pin of one mosfet would bear 35A and the other might not, though they are rated 74A.:-(