his paper introduces a novel algorithm, called Echidna, which solves multi-valued consensus, as well as two different approaches built upon it to achieve efficient State Machine Replication (SMR) with very low latency, Sphinx and Cerebrus, making them specially suited for decentralized payment systems. Sphinx has a leader-based design with a single proposer, while Cerebrus employs a leaderless strategy with multiple proposers. Both designs were implemented and their performance was evaluated in comparison with state-of-the-art decentralized solutions. Results showed that Sphinx and Cerebrus not only deliver comparable performance to them in terms of latency, but also present significant robustness under faults, while achieving enough scalability to handle typical usage levels of centralized payment systems.