… after all, you might expect a mature technology should no longer produce bad surprises.
One reason is that virtually every hydro plant has somewhat different rated data, accordingly also new tailor-made hydraulic profiles. Another reason is that even today the dynamic behaviour of the finished prototype cannot be predicted with certainty, not even after careful model testing. Some problems depend significantly from properties of the water passages in the plant, thus even identical machines in the same plant may behave differently.
Consider also the relative freedom of the designers who, in response to cost pressure and specification, constantly strive to find new detail solutions. Another factor is the complexity of problems that must not be underestimated. Researchers may boast glorious pioneering numerical simulations for various problems, but in practice there cannot be a scientific study for every aspect of every project.
New problems are also, in a way, generated by the purchasers of equipment who, following requirements of network regulation, tend to expand the desired range of operation more and more. Also the number of transient phases of operation like start and stop tends to increase. As a consequence, the machines must operate more often and longer in unfavourable conditions whose effects are not well understood so far.