BU87 Residual Mis-Pruning and the Co-Admissive Repository under B_U Short-Window Utility, Retainable Residual Structures, Same Settlement Surface, Grammar Consistency, and ReAdmission of Weak Persistent Morphisms

BU87 develops the theory of residual mis-pruning and the co-admissive repository under the B_U framework. Its central claim is that residuals are not automatically waste. Backward-mirror systems often treat residue, friction, delay, weak signal, local inefficiency, low-frequency recurrence, or task-window mismatch as harmful noise to be removed. BU87 argues that this cleaning impulse can destroy retainable structures before their settlement window opens. Residuals must therefore be classified, not blindly eliminated. The paper establishes a threefold residual classification. The first class is harmful bias: structures that continuously introduce semantic drift, task drift, constraint drift, or settlement drift. These require removal or strict isolation. The second class is removable friction: local cost or inefficiency with no observable recursive trace, cross-window relevance, or future re-admission potential. These may be compressed, pruned, or archived. The third class is retainable residual structure: weak, incomplete, low-intensity, or currently non-useful material that may carry weak persistent morphisms across larger time windows, broader task domains, or higher-order retained chains. This third class is the decisive object of BU87. BU87 defines residual mis-pruning as the premature removal of retainable residual structures under short-window optimization. A residual may lower immediate utility while supporting long-cycle continuity. In formal terms, a structure may satisfy J_{\tau}(\mu_w)0 in a larger evolutionary window. This mismatch produces what the paper calls the genius error: an effective higher-order path appears locally inefficient and is removed before it can become a standing channel. The consequence is closure of possible future retained chains. To prevent this failure, BU87 introduces the co-admissive repository. The repository is not a garbage bin. It is an observation, isolation, recombination, and re-admission zone for residual structures that have not yet obtained task standing but may carry future connective value. Residuals entering this repository do not receive immediate standing. They are preserved under classification, monitored for recurrence, tested for weak connectivity, and re-admitted only when boundary, grammar, feedback, and settlement conditions become compatible. The paper emphasizes that a repository works only under the same settlement surface and consistent grammar. Without a shared settlement surface, residuals remain a pile of unrelated fragments. Without grammar consistency, weak signals cannot recognize one another. When residuals share object schemas, boundary rules, state memory, feedback traces, friction-cost accounting, and re-admission conditions, rare elements can naturally gather into minimal structures. The paper therefore locks the chain: residual stratification, co-admissive repository, shared settlement surface, grammar consistency, weak persistent morphism, and new standing. BU87 extends the B_U sequence by showing how weak signals, failed attempts, edge data, immature fragments, recursive residues, and locally inefficient structures can become candidates for future standing. Its final judgment is that short-term optimization cannot directly adjudicate long-cycle evolutionary channels. Residuals must first pass through stratification, repository observation, and same-surface settlement before they are removed, retained, or re-admitted.