"Should" is not a good engineering word.
Yeah, you can put a diode to shunt kickback current.
Where "should" that current go, you may think you
know. But it's sharp and your model of the current loop
and the voltages around it, maybe not so accurate in
the time domain as your conception of it. Series R,
series L, high dI/dt, jack the ground and latch up a part
somewhere else?
So no more yammering about "should". Probe it, prove it,
fix it better next time, repeat.