Replacing a silicon switch with a silicon carbide or gallium nitride one looks like a drop-in improvement on paper: lower losses, higher frequency, smaller magnetics. The first prototype usually proves otherwise, and the fault is rarely the device.
Fast devices turn parasitics into design elements. Switching a few hundred volts in a handful of nanoseconds produces slew rates of tens of volts per nanosecond, and at that rate a few nanohenries of inductance in the gate loop becomes a resonant circuit with the gate capacitance. The results are familiar from support calls: ringing on the gate, false turn-on of the complementary device, erratic behaviour that changes when a probe is attached, and occasionally a device destroyed by a gate voltage nobody commanded.
The fixes are geometric rather than electronic. Keep the gate loop physically tiny, with the driver next to the device and the return routed directly beneath the drive trace so the loop encloses almost no area. Use the Kelvin source connection if the package offers one, so that current in the power path does not appear as voltage across the gate. Keep the power commutation loop short for the same reason, and resist the temptation to add gate resistance to calm things down — it works, at the cost of the switching losses that justified the device.
Then measure properly. A standard oscilloscope probe with a long ground lead will show you a ringing waveform that exists mostly in the probe. A short coaxial or active differential measurement at the device terminals tells you what the gate actually sees, which is the only waveform that decides whether the design survives.