Mirror Observerhood Lab II: Channel-Decomposed Reliability and the Limits of Reliability Tracking Alone

This release forms part of the Computational Observerhood Labs of Mirror Programme, Volume I: Observerhood. Lab II tests whether decomposing reliability into self-state, sensor-channel and world-map components is sufficient to improve viability under channel-specific perturbation. It follows Mirror Observerhood Lab I, which showed that scalar self-model reliability can help under false self-location but becomes brittle under generic sensor degradation. Across 5,760 deterministic grid-world episodes, four architectures are compared: a predictor-only agent, a self-model agent without reliability tracking, a scalar Mirror reliability agent and a decomposed Mirror reliability agent. The result is deliberately mixed. Channel decomposition improves attribution and some local properties, but it does not dominate scalar reliability or simpler self-model baselines. The central finding is negative but constructive: decomposed reliability alone is not enough. Reliability becomes observer-relevant only when channel-specific diagnosis is coupled to action, repair or commitment policies that improve viability after cost. The release includes a standalone paper and a reproducibility package containing the Python implementation, fixed-seed outputs, summary data, figures, requirements, citation metadata and licences.