I feel I'm so close and yet so far from understanding the control logic of shunt regulators. I will use as a reference "Sequential Switching Shunt Regulator for Satellite Power Control System" by A. Kamel et al., but there are several similar others.
I get how they detect a change in the load demands by sensing small changes in the main bus voltage, amplifying them and choosing the operating domain of the system. If there's an excess voltage then the batteries are charged, if there's even more then some energy is dissipated. On the other hand, if the voltage drops too low then the batteries come to help.
The question I cannot solve comes now: each error value corresponds to a predefined number of connected strings (you can see it in Kamel et al., table 2, I cannot add more links to the question). If for example a voltage excess is detected then some of the strings are shunted in order to limit the power. But by doing so, if I'm not mistaken, the voltage of the main bus would come down, and the error voltage with it. The system would think "with an error value this low I need to have more strings activated!" so more strings would connect, and you would get a voltage excess again. The system would, as I see it, be constantly jumping from one domain to another. Is that what would happen? Does that make sense? Or maybe I'm making a wrong assumption somewhere in my reasoning?