Repair storage and spatial contracts; defer optional imports and verify real services

[docxology]

Description

Correct DATA storage and spatial-index contracts: PostgreSQL generic values now support retrieval/deletion, IDs no longer collide within a second, writes reject replacement, and geospatial retrieval preserves CRS and quoted custom geometry names. H3/R-tree queries preserve duplicate/string observation labels and isolate indexed data from caller mutation.

Type of Change

• Bug fix
• Documentation update
• Refactoring

Modules Modified

DATA, SPACE, ANT, TEST, INTRA, root documentation and native manuscript rendering.

Related Issues

Addresses required readiness issues #61–68; final candidate evidence is accepted. Same-SHA main publication checks and fresh independent main review are now accepted. Unmet hardware, licensed-data, configured-advisor and release requirements remain open.

Changes Made

• Add real disposable PostGIS, Redis and MinIO/S3-compatible CI operations with tamper rejection, source/lock/runtime identities and retained failed outcomes.
• Defer optional SRAI imports during default H3 dispatch, preserving actual dependency failures on explicit admission; isolate ANT's 100-agent memory measurement in an owned child with the unchanged 200 MB cap.
• Declare GeoAlchemy2 in the PostgreSQL extra and renew generated README/AGENTS documentation.
• Remove forced manuscript section breaks that leave short tails on mostly empty pages; correct stale backlog and scientific-review criteria.

Testing

• DATA owning-module gate: 487 cases passed.
• Focused SQL/index/workflow tests passed; fresh independent source/infrastructure review completed and findings corrected.
• Strict documentation, Ruff lint and changed-file formatting, workflow lint and full-history plus final-delta secret scans passed.
• SPACE unit: 797 cases; ANT owning module: 179 cases; native SPM unit with documented macOS compiler adapter: 410 cases. Fresh independent source/adversarial review passed.
• Final candidate real PostGIS/Redis/MinIO receipts independently accepted, including tamper rejection and quoted projected geometry queries.
• Clean fea9f95706e618b61ff9b2cd1e955ef84eae007e: all eleven local producer groups passed; unit 45/45 modules, 10,883 cases.
• Native XeLaTeX manuscript: 30 pages, ROOT 144 and native layout seven cases passed; all four figures and 30 pages inspected. Whole-output snapshot: 2,038 files with matching before/copy/after hashes.
• All 30 PR checks green; full hosted archive and 46 coverage dispositions audited. TEST: 2,975/3,383 lines, 87.9397%, above unchanged 85%.
• Python 3.11/3.12: all 45 wheel builds/base imports and 20 operation profiles per interpreter independently verified.
• GitNexus at final revision, all 4,275 covered files match; generated signposts pass after indexing. Index omissions remain UNKNOWN.
• Final fresh independent aggregate, scientific and custody review passed for exact candidate fea9f95706e618b61ff9b2cd1e955ef84eae007e; all 32 raw candidate/supplementary archives accepted.

Documentation

• Generated README/AGENTS updated
• Storage contracts and issue reconciliation updated
• CHANGELOG updated

Checklist

• Source diffs reviewed
• Fresh independent shared-infrastructure/security review
• Dependencies documented
• All final source-bound local and hosted gates independently accepted before main integration

SQLite routing tests and patched PostGIS-writer tests are distinct from the new real-service CI lane. S3-compatible MinIO checks do not claim live AWS account acceptance. Native generation/rendering and coordinator inspection are renewed at the final candidate; independent aggregate/scientific review passed. Hosted producer uses its seven standard groups; the local full-validation record retains eleven.

The first local full run is retained as a failure: unprofiled macOS PyTensor linking failed, and ANT exceeded the memory cap in the shared pytest process. The original ANT RSS failure cause remains unproven. Final acceptance renews all groups after the reviewed changes and explicit native compiler setup; no deadlines, memory caps, coverage floors or scientific tolerances were increased.

Published to main by ordinary fast-forward at fea9f95706e618b61ff9b2cd1e955ef84eae007e; local, live remote and API identities agree. All 22 main CI jobs and separate main Windows/portability/GNN/format workflows passed. Fresh independent main review accepts all 32 raw archives and 46 measured coverage dispositions; all 45 unchanged module floors met. Actual main TEST: 2,976/3,383 lines (87.9693%), 1,976 cases, unchanged 85% floor. Required readiness issues #61–68 are fulfilled; six external/historical-cause issues remain open. Rollback ancestor: 42d2964b4a3919f3865b2b41a69acbe100b97c7c.