Limit Persistence: Margin–Mass Profiles, Full Histories, and Exact Hazard Cost
For compact spaces of compatible histories endowed with an execution law, we determine when viability at every finite horizon produces a positive mass of infinite histories and what cost accumulates at successive losses of margin. A single profile separates three thresholds: geometric possibility, accessibility from the initial law and positive execution mass. The full mass profile satisfies an exact factorization into initial survival and conditional hazards. Summability of these hazards preserves a positive mass of full histories whenever survival has positive mass at each finite horizon. The step-loss moment and the mass of the threshold layer yield a sharp interpolation law with an optimal constant. For a prescribed hazard, we determine the exact infimum of the corresponding moment cost and quantify the excess cost of an actual loss. A uniform tail transports the entire mass profile and its quantiles through limits of approximating families. Entropic tilting gives the minimal information cost of enforcing a prescribed persistence probability and the exact unresolved cost remaining after a finite prefix has been observed. In the Feshbach–Schur realization, the two positive components of the full Hermitian Gram matrix produce a quadratic hazard on the active fibers, while the graph lift accounts for preparation mass before conditioning. A telescope of overlapping frames gives the strong limit of the Gram matrix, an exact noncollapse criterion and a dark tail governing norm convergence; a separate criterion closes the phase. For signed pairs of legs, the minimal cover operator determines the full signature and a global norm trichotomy. Common domination of fresh layers extends to the complete tail, and in one dimension the tail coefficient is the exact weighted average of the layer coefficients.
Authors
- Maciej Stachowiak
- Paweł Nowak (ORCID: https://orcid.org/0009-0000-5562-1128)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-15
- DOI
- https://doi.org/10.5281/zenodo.24108267
- Primary Topic
- Probability and Statistical Research
- Type
- preprint