Corrosion fatigue is a typical failure mode of metallic materials subjected to the combined effects of cyclic loading and corrosive environments. It is widely observed in critical fields such as nuclear power, marine engineering, aerospace, and energy equipment, and directly affects the service safety and life assessment of engineering components. With the advancement of advanced energy systems operating in extreme environments such as deep space, deep sea, and deep earth, materials increasingly experience severe environmental-mechanical coupling damage. Among these environments, high-temperature pressurized water, liquid lead-bismuth, and marine conditions represent typical corrosive systems. Therefore, understanding and predicting the corrosion fatigue behavior of metallic materials under these conditions is of considerable importance. This paper reviews recent research progress on corrosion fatigue experimental techniques, damage mechanisms, and prediction models for metallic materials in the three representative corrosive environments mentioned above. Regarding experimental techniques, particular attention is given to the development of fatigue testing devices capable of simulating service environments, as well as in situ monitoring methods for specimen strain/displacement and crack length. In terms of damage mechanisms, the competition and synergistic interactions among several mechanisms are discussed, including stress concentration at corrosion pits, rupture of protective films and slip dissolution, hydrogen ingress and hydrogen-induced damage, and reductions in surface energy. For prediction models, the evolution from traditional empirical models, such as the Basquin and Coffin-Manson models, to data-driven machine learning approaches is summarized. The limitations of current models in terms of engineering applicability and integration of physical mechanisms are also highlighted. Furthermore, this paper discusses major challenges in the field, including the lack of experimental techniques for emerging extreme environments, insufficient understanding of multimechanism coupled damage theories, and the absence of high-precision life prediction models under small-sample conditions. Future research directions are proposed, including the development of cross-scale in situ characterization techniques, the integration of physical mechanisms with machine learning methods, and the advancement of design and evaluation systems for materials resistant to corrosion fatigue.