wouldn't supprise me if a typical waterwheel has a 2 or 3:1 ratio of unloaded to loaded rpm.
Pelton wheels have a 2:1 ratio between unloaded and optimally loaded.
A jet of water is reversed in the spoon as it pushes it. Optimally loaded,
the rotor moves at half the speed of the jet and the water leaving just
"falls down". Unloaded, the rotor moves at the speed of the jet.
Other wheel types may have other characteristics, depending on geometry.
But figure that they do something similar: Loaded, the water leaves with
negligible energy left in it, and the rotor was moving slowly in a way that
cushioned its exit. Unloaded, the water leaves with all the energy in it
that it had to start with, which means it's moving faster, so the rotor is
also moving along a bunch faster.
2:1 is a bunch of difference for frequency sensitive devices (though higher
frequency is generally better than lower, so it's not "saturate the transformer
cores and start a fire" nasty).
Trying to use a dump load to regulate the speed means keeping it fully
loaded all the time. That means the load is disappating as much energy
when the load is tiny as the full power of the mill. Think of that many
kilowatts of space heaters and you have an idea of how much power
your dump load must dissipate.
If your load isn't inherently adjustable to match the need, it seems to me
it's better regulate the incoming water with feedback to keep the rotor
at a particular speed, use some kind of transmission, or use a voltage
and frequency converting device and let the rotor speed and generator
voltage be "wild".