The objective of this paper is to investigate the cause of several unexpected high amplitude oscillations that occurred in the surge tank water level of a real hydropower plant during secondary load-frequency control (LFC) maneuvers, after the replacement of the turbine runner, and to propose solutions that allow the power plant to continue providing secondary LFC in a safe and reliable manner. For this purpose, a simulation model has been developed and calibrated from data gathered during several on-site tests. Two different solutions are proposed in order to cope with the observed problem: using a state-dependent load change rate limiter or modifying the hydro turbine governor gains; the turbine governor remains the same as before the runner replacement. The proposed solutions are tested against a set of realistic secondary LFC signals by means of simulations and compared to each other as a function of the probability that the surge tank water level descends below a minimum safe level and the quality of the secondary LFC response. The results presented in the paper demonstrate the validity of the methodology proposed to determine the state-dependent ramp limit, as well as its effectiveness to prevent the surge tank drawdown and to provide clear insight into the trade-off between response quality and power plant safety.