Lack of fusion (LOF) defects can significantly affect the mechanical properties of components manufactured by laser powder bed fusion (LPBF). The layer-wise evolution of LOF defects is complex and not yet thoroughly understood. This work explores the spatiotemporal variations of LOF defects under various conditions using in-situ monitoring, ex-situ surface topography and μ-CT porosity characterization, and high-fidelity multi-physics numerical simulation. The results show that LOF defects could exhibit typical self-healing characteristics for over five printing layers. The specific self-healing behaviors depend on the initial sizes of the defects and the LPBF process conditions. The evolution of LOF defects can be detected by in-situ monitoring using light intensity. However, the in-situ monitoring may miss detecting LOF defects buried below the healed layers, which were alternatively observed via μ-CT. For the defective area with a depth of 150 μm, the relative density increased from 61.7% to 95.7% for the first to the fifth printing layer. The optimization of process parameters demonstrated that the application of a 45° scanning angle could significantly enhance surface flatness and repair internal pores to a minimum of 36.0 μm. The findings highlight the ability of in-situ monitoring in detecting LOF defects and the potential of the controlled printing process to accelerate defect repair. These outcomes offer valuable insights for the industrial applications of in-situ monitoring and control during LPBF.