motor = generator (in reverse). sorry for the confusion.
The energy stored in the winding ½LI² will be converted to energy stored in the capacitor ½CV² minus resistive losses incurred while charging the capacitor. The rectifier diode will limit ringing but the resistance of the winding is probably very low so, in its current configuration, the capacitor surge current charging the capacitor is effectively the same amount of current flowing in the winding. At 40A, the capacitor will charge in about 1 millisecond. When the IGBT switches back on, it will discharge through the load in the same amount of time.
Electrolytic capacitors don't stand up to such treatment very well.
I somehow don't believe the 2A ripple current measurement can be accurate. A power resistor is usually put in series with the snubber capacitor to limit these surge currents and damp oscillations in whatever LC tank is present. Even with one in place, you may need more ripple current capacity than a reasonably sized electrolytic can deliver. I'd try four 100µF 400V caps wired 2x2 to provide the equivalent of an 800V 100µF with double the ripple current capacity if I had them but I still have my doubts.
A smaller setup here involving a motor less than 1/10th that size required a 50W 50Ω wirewound resistor and a 400V 10µF film capacitor to run at a relatively low 140V with a sub-1KHz switching frequency. The sub-1KHz switching made it obnoxiously loud.
I think 100µF of capacitance is probably about right for snubbing but a basic 400V 100µF electrolytic won't cut it. Think hundreds of Watts of resistance and an enormous 800V film capacitor if you're taking that path.
An alternate method used here: charge a 10,000uF 400V of electrolytics with the generator, use the IGBT/heater to discharge it at 4.88KHz, and place a much smaller snubber around the IGBT alone. The snubber will no longer be absorbing the energy stored in the collapsing field of the generator winding, just the much smaller field from the inductance that's present in the heater. With 10,000µF, ½LI² won't add up to much V in ½CV².