When you suppress an inductive spike with an over down capacitor, you increase the Ic squared *(ESR and Rdson) in the active devices. Null your conductor traces 1nH/mm when you have very fast switching times. You can reduce your dead time to sub microseconds and just allow for thermal delays from PTC effects of RdsOn. As long as the spikes are well below Vds Max and the slew rates with a tiny loop area will minimize your EMI radiation.
Keep in mind, you will always have an RLC spike response, including your 10:1 probe Gnd length and probe capacitance. This will help you understand what you were seeing.