I agree with you there, that carbon buildup is one main cause of the valves jamming, but is it probable that another cause is that someone has not calibrated the feedback position to RC valve actual position? So when the actuator drives to the end, the feedback says it's not in the correct position, starts building up the torque (and current) and eventually pops?
I've thought about fuses, but could a fuses give us the accuracy required? The theory behind it sounds good, as the I2t characterisitcs can be calcultaed but I don't know if a manufacturer would put that much precision into it considering they are designed for significant overcurrent.
CBs or a TOL type device I think is a good way to go, but like you I haven't put much more thought into it.
Good to see someone else has these thoughts simmering in the background