The observation

A “noisy” 3.3 V rail showed 400 mV of ringing at every logic edge. Before touching the circuit, I swapped the probe’s alligator ground lead for the ground spring. The ringing dropped to 70 mV.

Why

The alligator lead adds ~15 cm of wire - roughly 150 nH - in series with the probe ground. With the probe’s input capacitance that forms a resonant tank, and every fast edge rings it. Most of what the scope showed was the probe’s own resonance, not the circuit.

The habit

Measure fast edges with the shortest possible ground return, and be suspicious of any ringing whose frequency doesn’t change when the circuit does.