The debate between unified and partitioned consensus systems in distributed systems centers on their impact on latency. Unified consensus systems have a single leader region, which can perform linearizable reads across the servers without communication with other regions or waiting for clock skew uncertainty windows to close. In contrast, partitioned-consensus systems require multiple leaders per partition, leading to additional latency costs for multi-partition transactions and potentially higher long tail latency due to two-phase commit protocols. Furthermore, partitioned-consensus systems need clock synchronization to achieve low latency snapshot reads, which can result in serializability violations if the maximum clock skew bound assumption is not met. The disadvantages of partitioned-consensus systems relative to unified-consensus systems become more pronounced as applications scale, making unified consensus a safer option for many workloads.