ADR: Track high-complexity Rust module refactors (no mass rewrite)

Status

Accepted (2026-07-16) — tracking decision; implementation deferred to the sequenced plan.

Context

qlty smells --all (2026-07-15 audit) flagged large lib.rs monoliths in portable crates (templiqx-core, templiqx-application, templiqx-docx-v5, and secondary crates). Phase 0 of the smell plan landed: exclude worktrees and generated SDK sources so duplication noise no longer dominates the signal.

Mass refactors during concurrent feature work (SDK pilots, report-engine gaps) risk behavior drift on CRM3 grounding and fail-closed validation.

Decision

1. Do not perform large complexity rewrites as part of readiness cleanup.
2. Track splits in docs/plans/2026-07-15-002-chore-qlty-smells-refactor-optimize-plan.md with inventory in docs/audits/2026-07-15-qlty-smells-findings.md.
3. Keep qlty smells advisory ([smells] mode = "comment"). Gate on qlty check --level=low and Clippy -D warnings only.
4. Execute Phases 1–4 one crate per PR, behavior-preserving moves first, complexity helpers only with golden/conformance coverage.
5. Suppress remaining noise via .qlty/qlty.toml excludes (generated SDKs, worktrees, scratchpad/, artifacts/) rather than silencing real product complexity.

Priority backlog (from plan)

Non-goals

• Failing CI on smell complexity thresholds before Phase 1–3 land.
• Rewriting DOCX dialect claims or relaxing templiqx/v1alpha1 validation.
• Refactoring generated OpenAPI client sources.