fix(drift): discount CALLS-layer tangle by unresolved_ratio (#244)

[zaebee]

Summary

Closes #244 (colleague-review note 1 from PR #241). PatternFingerprint.tangle_ratio now fades the CALLS layer by the confidence discount clip(1 − unresolved_ratio) before the max():

calls_discount = max(0.0, min(1.0 - self.unresolved_ratio, 1.0))
return max(tangle_mass(self.t_imports), calls_discount * tangle_mass(self.t_calls))

t_calls is a census over resolved intra-domain CALLS only, so a domain whose calls are mostly raw/unresolved — plus one resolved mutual triple — could read a high calls-tangle (tangle_mass ≈ 0.67 from a single 201) and trip the hard hygiene gate on thin evidence. This mirrors the existing drift-distance treatment (_clip_discount / _score_v2), which already fades the calls layer for exactly this reason. The IMPORTS layer (always resolved) is never discounted.

Why it's safe

The discount can only lower calls-tangle, so it can never push a green domain red — the max: 0.25 threshold stays valid and no re-measurement is needed.

• Self-drift verified: cgis drift on the self-graph → 0 tangle gate_failed (unchanged).
• Full suite green (1060), ruff + mypy strict clean, doc-coverage 99.6%.

Tests

• calls-tangle discounted by unresolved_ratio (mesh + 0.5 unresolved → 0.5);
• fully-unresolved calls fade to 0 (cannot trip the gate);
• IMPORTS layer not discounted (regression guard);
• fully-resolved calls unchanged (regression guard).

Residual (deliberately not in scope)

This fixes the mostly-unresolved case (issue option 1). It does not address a domain with few but fully-resolved mutual calls (unresolved_ratio = 0 → discount = 1 → full calls-tangle from one triple) — that's issue option 2 (a minimum-triad-count floor). Left out per YAGNI: it needs its own magic-number calibration, the common case is the unresolved one, and self-drift is safe regardless. Easy to add later if a real domain hits it.

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

🤖 Generated with Claude Code

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[gemini-code-assist]

Code Review

This pull request implements a confidence discount for the CALLS layer in the tangle_ratio calculation, addressing issue #244. The CALLS layer is now faded by 1 - unresolved_ratio (clipped between 0.0 and 1.0) to prevent sparse or mostly-unresolved call graphs from over-reporting calls-tangle. Unit tests have been added to verify this discounting behavior, ensuring that the IMPORTS layer remains unaffected and that the discount behaves correctly under various unresolved ratios. No review comments were provided, so there is no additional feedback to address.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.