Leakage as a Breakdown of Expected Feedback Pathways: A Dynamic, Multi‑Scale, and Observer‑Aware Framework for Classifying Stability, Failure, and Collapse in Complex Systems

Abstract This paper presents a dynamic, multi‑scale, and observer‑aware analytical framework for classifying system behavior in response to failure. The concept of “leakage” is redefined as a breakdown of expected feedback pathways. Unlike static typologies, this framework introduces transition rules between system states based on the accumulation and processing of leakages over time. It explicitly addresses the observability tension by distinguishing between the operational system (actual dynamics) and the observed system (what an observer can detect), showing that the same operational state can appear differently depending on the observer’s access. Finally, it specifies three analytical scales (micro – component, meso – subsystem, macro – whole system) and demonstrates how leakage can propagate across scales. A set of dynamic indicators, falsifiability conditions, and practical assessment probes are provided. Keywords: Leakage, system dynamics, observability tension, multi‑scale analysis, complex systems, feedback breakdown