Long-term, sustained physical activity is crucial for health, making it a holy grail for physical activity research. Researchers often refer to it as 'physical activity maintenance.' Yet, this concept lacks consensus, possibly due to the absence of an empirically testable and demarcated definition. This study aims to explore the boundary conditions of two operational definitions of physical activity maintenance (activity threshold and timescale) and examine how their formulation impacts the detection of sustained physical activity patterns. We conducted secondary analyses of 352 time series (length=182 days) of moderate-to-vigorous physical activity (MVPA) collected daily with Fitbit from participants in a weight loss intervention. We defined two activity thresholds based on participants’ baseline intentions and physical activity levels: a) 150 minutes of MVPA/week for participants below this threshold at baseline (subsample 1; n=101), and b) current level of MVPA/week for those already exceeding 150 minutes of MVPA/week at baseline (subsample 2; n=251). We used regression trees to examine changes in physical activity trajectories using different boundary definitions of timescales (7, 14, 28, and 56 days). Results showed that at finer-grained timescales (e.g., 7 days), physical activity exhibited repeated fluctuations, with sustained activity patterns observed in 27% and 5% of participants in subsamples 1 and 2 respectively. At coarser-grained timescales (e.g., 56 days), physical activity trajectories were more stable, with sustained activity percentages increasing to 71% and 26%. This study highlights how varying the boundary conditions for defining long-term sustained physical activity can lead to differing conclusions regarding a person's long-term activity level. Future work should more carefully study boundary condition specifications and their impact within the field.