Relational Reversibility: Structural Conditions for Recovery Capacity Generation in Adaptive Multi-Agent Systems

RT4 studies how interaction changes a declared system's ability to recover after disruption. It distinguishes recovery supplied by current support from capacity retained after earlier cooperative learning or repair, and separates autonomous, assisted, and preparation effects. Support paths and failure-transmission paths are treated as different objects. An exact finite-cascade benchmark isolates support loss, transmission removal, and the task cost of hub removal. A trust-connectivity score remains a candidate predictor rather than a general recovery probability. Unreproducible demonstrations and an undocumented collaboration analysis are retired; causal capacity-generation mechanisms and incremental predictive value require direct experiments.