Decomposing the Disappearance of Cold Zones into Two Mechanisms: A Public Analysis Protocol for Existing Data

Abstract An increase in the screening length erases spatial cold zones. The screening length can grow along two distinct routes — a rise in transport capacity, or a fall in local dissipation — and the two are observationally separable: the first is read from accessibility, the second from the relaxation curve of a disturbance. This paper registers a public analysis protocol for existing historical data; it reports no empirical results. The protocol fixes: a geometric operational definition of the cold zone that contains no absolute field-strength threshold, only one explicitly registered relative threshold; computational rules for the cold-zone fraction, the redundancy index and the effective number of independent flows, together with a parameter-free out-of-sample prediction derived from a substrate-sharing matrix; a two-mechanism coding procedure executed by hypothesis-blind coders on de-identified material; and an exhaustive ex ante registry of all analytic degrees of freedom. The principal result produced here is negative. A simulation of classifiability shows that, with an equivalence bound of ln 1.5, a median log-scale standard error of 0.15 yields a classifiable-period rate of 0.675 but a recall of only 0.570 for the dissipation-collapse type; recall falls to 0.301 and 0.131 at standard errors of 0.20 and 0.25. Two parallel precision gates are registered accordingly, together with a fixed reading rule: failure to observe the dissipation-collapse type may never be read as its absence. A second simulation prices the registered equivalence bound: a pure mixed case in which both sides change by a factor of 1.4 is misclassified as one of the two pure types in 11.3% of runs at a standard error of 0.15, rising to 36.3% if the bound is relaxed to ln 2. The same simulation overturns a clause the protocol had already registered — the magnitude-comparison gate intended to suppress such misclassification is logically redundant across the entire registered precision range and changes essentially nothing. This paper is an analysis plan and contains no empirical results; H1–H5 constitute an executable and auditable testing framework and must not be cited as verified findings. Because the data were already public when the protocol was written, every test registered here is a post hoc test and provides exploratory evidence only.