In recent years, Multi-Connectivity (MC) has emerged as a promising approach to enhancing the reliability and resilience of communication networks. The effectiveness of MC depends on the statistical dependencies between the links and heavily on the selection of an appropriate MC strategy. We introduce Packet Duplication (PD) and Load Balancing (LB) in this paper. Each of these strategies offers distinct advantages, but a comprehensive comparison under varying network conditions and dependencies is missing. To address this gap, we investigate an idealized diamond network element, applying two strategies. Analytical results are provided, highlighting the mathematical relationship between effective rate and reliability, considering worst-, best-, and independent and identically distributed (i.i.d.)-dependencies between links. Numerical simulations demonstrate the dichotomy of the optimal strategy under different network scenarios. Interestingly, LB outperforms PD for the worst-case. Additionally, we extend our analysis by deriving reliability bounds as functions of the effective rate for arbitrary network structures in both PD and LB.