The marine environment is complex and variable. To enhance the autonomous navigation and fault response capabilities of underactuated marine vehicle (UMV), this paper proposes a finite-time sideslip-tangent line-of-sight (LOS) guidance-based command filtering path-following control scheme. This scheme enables UMV to follow the desired path within a finite time, even under complex conditions such as unknown internal dynamics, time-varying large sideslip, and actuator failures. Firstly, a finite-time sideslip observer is constructed to rapidly and accurately estimate the time-varying large sideslip. Subsequently, a finite-time sideslip-tangent LOS guidance law is introduced, which not only improves guidance performance but also effectively avoids high-frequency oscillations caused by non-smooth guidance commands, thus preventing instability in UMV. Additionally, finite-time command filtered control technology is employed to reduce computational burden, and a filtering compensation scheme is proposed to minimize filtering errors. Finally, based on radial basis function neural networks and finite-time theory, an adaptive finite-time fault-tolerant path-following controller is designed, which enables both surge velocity and heading angle tracking errors converge to a small neighborhood around the origin within a finite time. Simulation experiments verify the effectiveness and superiority of the proposed scheme.