Well, IEC-International Electrotechnical Commission states that for voltages around 50V human body on a hand-to-hand pattern has an impedance of about 1500 Ohm or even less, that one significantly depending on actual body mass, skin conditions, contact area, applied voltage, frequency, etc.
Now the point: at 50 km/h at no-load conditions, or on a 1500 Ohm load with no significant difference at all, out of a dynamo without a built-in voltage protection you may have something like 50V that applied to your body might induce in turn something like 30 mA.
Something definitely unlikely to happen, I agree, but still possible … especially for grounded hub dynamos and spoilt head lamps with exposed/not well protected live parts.
From the point of view of a possible exposure hazard IEC worked out the below Current-vs-Time plot identifying four zones:
Zone 1: represents the limit for current perception estimated to be 0.5 mA
Zone 2: represents the danger threshold generally recognized still to have no dangerous physiological effects
Zone 3: it’s a sort of an “alert” area coming just before any possible atrial fibrillation, mainly characterized by yet reversible physiological effects such as muscular contraction (tetanization), difficult respiration and cardiac disturbances. As you can see 30 mA for one sec may already have some annoying effect.
Zone 4:is characterized by permanent effects, such as fibrillation, depending, beyond current and time, also on specific health conditions. Contact times as low as 10 ms may be lethal but fortunately these current values seem to be much higher than those ones actually supplied by a dynamo at 50 km/h !!
Here below an example of how, e.g. SHIMANO, implemented a surge protection in its switch SW-NX10; in this case ZD1/ZD2 zener protection is there not for "regulating" but for avoiding to have dangerous voltages on the "live wire" whilst the S1 switch is open, i.e. when the lights are off. In such a case indeed, in the event to have a voltage surge this would be routed through the bridge instead of through the cyclist in the always possible circumstance to have an accidental contact with the live parts of the circuit.