Blockchains promise to make online services more fault tolerant because they are replicated on a distributed system of nodes. Their nodes typically run different implementations of the same protocol across different geo-distributed regions, making the protocol supposedly tolerant to various failures including isolated crashes, transient failures, network partitions or attacks. Unfortunately, their fault tolerance has never been compared. In this paper, we provide the first fault tolerance comparison of blockchain systems. To this end, we introduce a novel sensitivity metric, interesting in its own right, as the responsiveness difference between a baseline environment and an adversarial environment. We inject various failures in controlled deployments of five modern blockchain systems, namely Algorand, Aptos, Avalanche, Redbelly and Solana. Our results show that (i) all blockchains except Redbelly are highly impacted by isolated failures, (ii) Avalanche and Redbelly benefit from the redundant information needed for Byzantine fault tolerance while others are hampered by it, and more dramatically (iii) Avalanche and Solana cannot recover from transient failures.