When you spin a motor connected to ODrive, the motor naturally generates an AC voltage proportional to its speed, and this will drive a current through the body diodes of the FETs, which act like a bridge rectifier and will charge up the capacitors. If, for some reason it draws a lot of current through this bridge rectifier, then it will cause the FETs to heat up, producing about 2W per amp as heat due to the 0.7V silicon diode voltage drop across each of the FETs when back-driven like this.
Also, the ODrive will turn on (you will see the green light come on) and it will potentially start running its startup actions. If the startup actions include motor or encoder calibration, you could be in trouble if it tried to do that with a fast-spinning motor. It would energise the coils at a random timing with respect to the spinning motor, which could produce a higher voltage than the motor’s EMF at that speed.
What was the mechanical situation that led to this? A runaway winch? A vehicle rolling downhill?
Is it possible that the motor spun significantly (more than 20%) faster than would normally be its maximum speed if you set a continuous torque demand at 56V input?
If so, then the voltage generated by the motor could be higher than the capacitors, FETs and gate drivers on the board can handle.
BTW: Does M1 still work?