Hands-on surgical training (HOST) for congenital heart surgery (CHS), utilizing silicone-molded models created from 3D-printing of patients' imaging data, was shown to improve surgical skills. However, the impact of repetition and frequency of repetition in retaining skills has not been previously investigated. We aimed to longitudinally evaluate the outcome for HOST on two example procedures of different technical difficulties with repeated attempts over a 15-week period. Five CHS trainees were prospectively recruited. Repair of coarctation of the aorta (CoA) and arterial switch operation (ASO) were selected as example procedures of relatively low and high technical difficulty. Procedural time and technical performance (using procedure-specific assessment tools by the participant, a peer-reviewer, and the proctor) were measured. Coarctation repair performance scores improved after the first repetition but remained unchanged at the follow-up session. Likewise, CoA procedural time showed an early reduction but then remained stable (mean [standard deviation]: 29[14] vs 25[15] vs 23[9] min at 0, 1, and 4 weeks). Conversely, ASO performance scores improved during the first repetitions, but decreased after a longer time delay (>9 weeks). Arterial switch operation procedural time showed modest improvements across simulations but significantly reduced from the first to the last attempt: 119[20] versus 106[28] min at 0 and 15 weeks,  = .049. Complex procedures require multiple HOST repetitions, without excessive time delay to maintain long-term skills improvement. Conversely, a single session may be planned for simple procedures to achieve satisfactory medium-term results. Importantly, a consistent reduction in procedural times was recorded, supporting increased surgical efficiency.