Understanding how students' learning approaches differ and appear to shift across academic years is central to improving academic outcomes and designing effective pedagogy. While much research assumes a progressive shift toward deeper strategies, this trajectory is neither universal nor guaranteed. This mixed-method study examined study approaches and learning behaviors among 220 Indian undergraduates across academic years using the R-SPQ-2F and qualitative accounts. Quantitative analyses (SEM, MANCOVA, Multiple Linear Regression, and Chi square tests) showed that deep learning was associated with consistent routines, high concentration, and effortful study method preferences, whereas surface learning was linked to male students, lower prior achievement, poor attendance, reliance on YouTube or AI tools, and lower GPA. Academic year did not significantly predict learning approaches, a phenomenon we term "learned stagnation," yet behavioral trends such as declining attendance, reduced satisfaction, and greater dependence on AI suggested a gradual drift toward surface strategies over time. Thematic analysis revealed challenges of autonomy, perceptions of grading fairness, digital distractions, and psychological fatigue, but also highlighted gains in maturity, independence, and self-directed learning. Findings call for structured pedagogical support and responsible technology use to foster deep learning and resilience. This study advances global scholarship by introducing learned stagnation and extending Biggs' 3P model to include structural and technological factors, offering a nuanced Indian perspective that challenges the assumption of universally deepening study approaches.