Background: Post-stroke walking impairment is a significant cause of chronic disability worldwide and often leads to loss of life roles for survivors and their caregivers. Walking impairment is traditionally classified into mild (>0.8 m/s), moderate (0.41–0.8 m/s), and severe (≤0.4 m/s), and those categorized as “severe” are more likely to be homebound and at greater risk of falls, fractures, and rehospitalization. In addition, there are minimal effective walking rehabilitation strategies currently available for this subgroup. Backward locomotor treadmill training (BLTT) is a novel and promising training approach that has been demonstrated to be safe and feasible across all levels of impairment; however, its benefits across baseline walking impairment levels (severe (≤0.4 m/s) vs. mild–moderate (>0.4 m/s)) have not been examined. Methods: Thirty-nine adults (>6 months post-stroke) underwent 6 days of BLTT (3×/week) over 2 weeks. Baseline and PRE to POST changes were measured during treadmill training and overground walking. Results: Individuals with baseline severe walking impairment were at a more significant functional disadvantage across all spatiotemporal walking measures at baseline and demonstrated fewer overall gains post-training. However, contrary to our working hypothesis, both groups experienced comparable increases in cadence, bilateral percent single support times, and step lengths. Conclusion: BLTT is well tolerated and beneficial across all walking impairment levels, and baseline walking speed (≤0.4 m/s) should serve as a covariate in the design of future walking rehabilitation trials.