Native self-analysis: triage structural hotspots and record revision-bound dispositions #76

The 0.6 native analyzer was exercised on its own exact release commit 09d497f16520b6dc20f5dd42077ff76774e1f5e1 with default policy devostasis.code-policy.v1/default-1. Source and recorded history coverage are COMPLETE; the source-proof bundle verifies offline (bundle ID sha256:8a4d6142cdb920c1d3c8e2bd7d28d37cf88083453ee01f56a6f057e6649418ad). It analyzed 103 Python files and 1387 named functions.

There are 82 local static triage findings: 56 decision-count, 19 function-span, 6 file-size and 1 static import-cycle warning. These are policy triggers, not 82 confirmed defects or a health score. Tests and exhaustive contract tables can legitimately exceed these local thresholds.

Representative production candidates (path, symbol, starting line, line span, body decision count):

• src/devostasis/workscope/planner.py, project, 58, 156, 85.
• src/devostasis/adapters/github.py, GitHubAdapter._collect_ci, 1300, 213, 73.
• src/devostasis/bundle.py, verify_members, 576, 143, 70.
• src/devostasis/codeanalysis/engine.py, evaluate, 108, 103, 62.
• src/devostasis/codeanalysis/bundle.py, admit, 34, 97, 59.
• src/devostasis/codeanalysis/git.py, collect, 96, 111, 49.

Follow-up acceptance:

1. Reproduce native analysis at the pinned commit, then classify representative source warnings with concrete maintainability evidence; keep intentional exhaustive tables separate.
2. For justified refactors, extract named admission/acquisition phases while preserving canonical bytes, availability gaps, source proof and historical replay. Add cases only for newly exposed behavioral boundaries.
3. Record a revision-bound disposition and rationale for deferred/intentional warnings. A moved finding or missing source is not accepted as improvement.
4. Inspect the static import cycle for actual initialization/maintenance consequences before asserting a runtime defect.

This is bounded review work, not an additional accepted research rule or an automatic change to Vitals/thresholds. Related: #72, #73, PR #75.