Conventional gait rehabilitation methods have the risk of alienating the patient due to their monotonous nature, thus negatively impacting the effectiveness of gait training. Modern technologies can help provide patients with better support, safety and immersive experience during training. However, physiotherapists cannot be required to master those technologies, nor to spend much time designing a more varied and engaging treatment for each patient. In this paper, we argue that adaptive gamified gait rehabilitation based on procedural content generation (PCG) can effectively support physiotherapists in achieving such customized outcomes. We propose a generic adaptation scheme to steer the generation of movement challenge levels based on player modeling and therapists' intervention. Our approach features two difficulty adjustment strategies: parameter progression schemes and integration of multiple therapy goals. These strategies are applicable to the personalization of a wide range of gait rehabilitation goals. We implemented this approach in a standalone prototype for supporting gait training with the RYSEN system, a three-dimensional overground body weight support system. From our assessment with physiotherapists, we conclude that our PCG-based adaptive method effectively assists therapists in (i) offering a broad diversity in gait exercises to a wide group of patients, and (ii) dynamically tailoring challenge levels for a variety of gait tasks.