[noahgift]

What

decide() in aprender-review-experiment caps a lane at Tripwire when its primary cell is devbox or unnamed (S-4), but recorded no reason for the cap. It now pushes one S-4: reason, so the decision log says why the lane dropped. A reason already present from evidence() is not duplicated.

Why

Follow-up to the review finding on #5393: a capped lane produced a mode change with an empty reason trail.

Falsification

FALSIFY-RXG-008 (ladder_tests.rs) asserts exactly one S-4: reason for vote and tie-breaker lanes. It goes red with the fix reverted.

Verification

• cargo test -p aprender-review-experiment --lib: 234 passed
• clippy -D warnings and cargo fmt --all -- --check: clean

Refs #4354
keep-open: #4354 stays open, this is one step of its work.

🤖 Generated with Claude Code

[noahgift]

seat 2, aprender-seat-s12 @9852b061a5 APPROVE (no blocking finding)

Read-only review of the one Opus-authored commit (trailer Claude Opus 5.5), 2 files: ladder.rs +6, ladder_tests.rs +11.

• Before the change, decide()'s S-4 cap path (devbox or unnamed primary, current > Tripwire) set mode to Tripwire and returned, without recording a reason when the report alone was already at or below Tripwire. Only capped() pushes "S-4:", and only when its own mode is above Tripwire. The patch adds the reason in decide() guarded by !reasons.any(starts_with "S-4:"), so the capped() case is not double-counted.
• d is already mut. Mode derives Debug, so the {current:?} and {:?} format args compile. The mode is read before it is overwritten, so the text reports the report-alone mode.
• Tests: the loop over Vote and TieBreaker on the devbox_h4_fail fixture (report alone = Shadow) asserts Tripwire plus exactly one "S-4:" reason. The devbox fixture at Vote asserts exactly one, which pins the dedupe where capped() already pushed. Both directions of the count are covered.
• Not run locally: no isolated build of the PR head here. CI shards (workspace-test-shard 1-3, x86-main) were still pending when I read them, so the tests' pass state rests on CI.

[github-actions]

§13.11 rung 1 — quorum shadow verdict

S13-SHADOW pr=5415 head=9852b061a561f0e475dc2aba0c0cffdf49cf070b verdict=REFUSE class=Q1 arm_rc=1

Shadow mode: this records a verdict and merges nothing. A refusal
to arm is not a block (§13 adds zero rows to §7) — the pull request is
exactly as green as it was.

[noahgift]

seat 1, aprender-seat-s11 @9852b061a5 APPROVE (no blocking finding)

Read the 2-file diff and ran cargo test -p aprender-review-experiment --lib ladder in a throwaway worktree at 9852b06 (removed after the run): 10 passed, 0 failed, including falsify_rxg_008. I did not open the C314 receipt.

• The early-return S-4 branch in decide() now records its reason. It only adds an "S-4:" reason when capped() did not already add one, so the reason appears exactly once. The mode result is unchanged: still Tripwire, same early return.
• Tests: Vote and TieBreaker lanes on a devbox report whose own verdict is Shadow are capped at Tripwire with exactly one "S-4:" reason. The plain devbox report on a Vote lane also shows exactly one S-4 reason, which covers the path where capped() already pushed it.

Non-blocking: when the report alone says Shadow, the capped lane moves up to Tripwire. That predates this PR, and the new reason text states it ("the report alone gives Shadow").