Background/Objectives: Falls and related injuries are a frequent and serious health problem in older persons. Among the various strategies, different forms of active physical training, in particular, have demonstrated success in reducing fall risk. A task-specific training approach is perturbation-based training of reactive balance. Performing this training modality on a perturbation treadmill, secured with a safety harness, is an innovative new approach facilitating task-specific training with unannounced perturbations in a safe environment. The aim of this study was to investigate the feasibility and effectiveness of this specific training in multimorbid older hospitalized patients with prefrailty and frailty. Methods: The trial was conducted as a prospective single-center, assessor-blinded randomized controlled trial. A total of 127 acute-care geriatric hospitalized patients were enrolled in a program either involving a minimum of 60 min perturbation-based treadmill training or treadmill training without perturbations on the identical device and for a comparable training period. Results: Participants were 81 ± 6 years old (64% female) with a baseline FRAIL Scale, SPPB, and MoCA scores of 3.5 ± 1.6, 8.3 ± 2.6, and 21 ± 5 points, respectively. The training was performed on six occasions with an average total training period of 89 min during a mean hospital stay of 17 ± 3 days. Between the baseline and up to 2 days after the last training, the Short Physical Performance Battery score, which was considered the primary endpoint, improved by 1.4 ± 2.1 points in the intervention group compared to 0.5 ± 1.7 in the control group, with a 0.9-point difference between the groups (p < 0.001). Conclusions: a relatively short training period of approximately 90 min on a perturbation treadmill led to a significant and clinically meaningful increase in the physical performance of frail and prefrail hospitalized geriatric patients. However, its effectiveness in reducing fall risk is yet to be proven in this population.