[ENG-25] Many noqa: C901 / PLR0912 / PLR0915 / PLR0913 suppressions (69 instances)

[kgdunn]

Tracked as ENG-25 in ENGINEERING_AUDIT.md.

Severity: Moderate

Problem

ruff's complexity checks are silenced in-line via # noqa at 69 sites across the codebase (C901, PLR0912, PLR0915, PLR0913). The presence of so many is a code-smell canary: these are exactly the functions most likely to grow SEC-NN-style bugs (they are complex enough that no single reader can hold them in their head). SEC-21 sub-items 1-9 all live inside functions that carry these suppressions.

Fix direction

Treat each noqa: C901 as a refactor candidate. Pick the worst offender per release cycle and split it. Track progress in a checklist.

Acceptance

• Document a target (e.g. "halve the suppression count by v2.0").
• Each release lowers the count.
• New code does not add noqa: C901 without a comment explaining why a refactor is deferred.

Notes

This is a long-running effort, not a single PR. A good way to chunk it is: one suppression removed per code-review cycle in the relevant module.

[kgdunn]

Progress (chip-away, per the "worst offender per release cycle" cadence):

• ENG-25: chip away screening-design complexity suppression #360 (v1.24.30): split _select_screening_design (experiments/strategy/engine.py) into four focused helpers and dropped a dead reasoning list. Removed its # noqa: C901, PLR0912, PLR0915. Behaviour-preserving (75 strategy tests pass identically).

Suppression count: 68 -> 67. Issue stays open as the ongoing tracker.

---

Generated by Claude Code