This paper advances Nervous System Intelligence (NSI) Theory by formalising recovery dynamics as the primary basis for inferring whether change is biologically durable. Building on the architectural constraints defined in NSI Theory I, the paper introduces a recovery-first inference framework that distinguishes integrated recovery, compensatory stability, debt-accumulating recovery, and fragmented recovery following identical perturbation. Recovery trajectories are treated as lawful system responses arising from underlying biological readiness rather than from differences in load, intervention, or effort. The paper demonstrates how superficially successful change may stabilise temporarily while remaining biologically unstable, accrue recovery debt, or collapse over time. NSI Theory II establishes recovery dynamics as a necessary analytical layer for distinguishing durable reorganisation from transient adaptation, providing falsifiable conceptual criteria for evaluating whether change is capable of holding. This work constitutes Theory Paper II in a three-paper theoretical series. It is published as a preprint to establish conceptual foundations and invite scholarly dialogue prior to empirical operationalisation and peer-reviewed submission. Author Terminology Update as of 31 January 2026 This paper was published under the term Nervous System Intelligence Theory. The theory is now situated within the broader framework of Human Operating System Architecture (HOSA), reflecting expanded scope and interdisciplinary applicability. The underlying mechanisms, principles and claims remain unchanged.