In subduction zones, fluids are often invoked to explain slip processes on the megathrust, from great earthquakes to slow-slip events and tectonic tremors. However, it is unclear how the transient evolution of pore-fluid is controlled by depth-dependent variations in hydraulic properties over a broad range of timescales concomitant with the full spectrum of seismic and aseismic slip. In this study, we leverage a newly-developed fully dynamic hydro-mechanical earthquake cycle modeling framework to simulate fluid-driven seismic and aseismic fault slip. By assimilating geological, geophysical, and laboratory data in a physics-based model of fault dynamics, we investigate the role of hydraulic properties on-fault in controlling the predominant slip mode along the subduction megathrust. Results indicate that fluid-driven shear cracks nucleate due to a competing mechanism between the compaction of pores and the dynamic self-pressurization of fluids inside the megathrust, whereas the subsequent propagation of dynamic ruptures is self-sustained by solitary pore-pressure waves. While models with uniform hydraulic properties yield to regular seismic cycles of complete megathrust ruptures, a depth-varying fault permeability leads to the emergence of complex aperiodic sequences characterized by partial and complete ruptures, aftershocks, and transient aseismic slip. Further parameter analysis shows that the slip response on-fault primarily depends on fault permeability and porosity, which in turn control the poroelastic compaction, the storage capacity, and the hydraulic diffusion length. Four slip response patterns are revealed by the parameter space, including seismic events, slow-slip events, oscillatory decay with time, and stable aseismic creep. Our findings provide new insights into the interplay between pore-fluid, mechanical, and fault slip processes, and suggest that solid-fluid interactions and the permeability architecture play a key role in controlling the predominant slip mode on subduction megathrusts.