Paper-lane evidence: native-fault receipt on the order-contract engine; first 1x leverage-ladder attempt (needs_attention)

[seathatflowsinourveins]

What

This PR adds two paper-lane records from the WSL host's dedicated Alpaca paper account on 2026-09-24. Neither changes a gate status or any engine code.

1. Native-fault receipt on the order-contract engine (native-fault-behaviour stays established)

Main's native-faults README said the order-contract boundary (#198) had "local synthetic test coverage only; no native run has exercised it". The 18:58Z run is that native run.

	14:39Z run (the flip commit's receipt)	18:58Z run (new receipt.json)
Engine	before #198	#198 head 81cbb06; every bound hash equals main at 3b7ae71
C01 / C02 / C05 / C04	200 / 204 / 204 / 422 then 404	200 / 204 / 204 / 422 then 404
Cleanup	flat, cash delta 0.00	flat, cash delta 0.00
Independent observation	independent-observation-20260924.json	independent-observation-20260924t185811.json

• The 14:39Z receipt is kept byte-for-byte as receipt-20260924t143905.json (sha256 e3cec5a6…).
• The gate note gains one dated sentence.
• Evidence class: native_paper for each case. The observation is independent of the harness and engine code, not of the broker.

2. First 1x leverage-ladder attempt, retained as a failure (leverage-ladder-1x stays not_established)

Run. Trial ladder-1x-20260924a used config-leverage-1x.json from a frozen archive of main 3b7ae71.

• It ended needs_attention after 190 s of 300.
• The account was flat and reconciled: 4 orders, all filled, net +1.20 USD.
• Peak achieved leverage was 0.151, with 0 s above 0.5x, so the rung's exposure condition was not met either.

Why it stopped: a false-positive fill-precision refusal (order_cumulative_fill_precision_requires_reconciliation).

• The INTC sell filled 6 @ 126.33, then 1 @ 126.31.
• Alpaca's six-decimal average 126.327143 made the adapter derive 126.310001 for the last share, which is off the cent grid.
• The broker's own /v2/account/activities/FILL readback shows the on-grid executions.
• This is the defect class Mover mode's first Alpaca paper runs on the Mac (trials A and C) and the partial-fill precision fix #205 fixes. The sequence here makes a concrete regression case: 6 @ 126.33, then 7 @ an average of 126.327143, with an expected last price of 126.31.

Hypotheses, recorded as such:

• Regime availability. The leverage ceiling was 0 for 98 of the 189 s, because the regime was unavailable whenever a benchmark lacked features. IWM had 25 of the 37 dropped crossed quotes.
• Rung reachability. Reaching 0.5x needs at least six concurrent positions of about $880 under the $1,000 per-order cap, and portfolio_rotation exited within 6–35 s.

Host setup. The promotion-gate runtime scheduled_trial.sh expects was missing on this host. It was installed the same day from blueprints/us-equities/data/requirements.lock using uv pip sync --require-hashes.

Checks

• validate.py, validate_catalogs.py, landscape.py --root ., evidence_manifest.py --check (5314 files), verdict_review_gate.py --base origin/main (no verdict rows changed) and trading_gates.py --check all pass.
• tests.test_native_faults_min and tests.test_adaptive_paper_runner pass on the pinned runtime: 150 OK.
• tests.test_trading_gates, test_evidence_manifest and test_validate pass: 88 OK.

Not covered

• The 18:58Z run's exit code was not retained. By harness design, native_faults_passed is exit 0.
• The in-progress corporate-action guard branch will change runner.py again. After it merges, this receipt binds the older runner until a new run is recorded, as the README states.
• No other ladder rung was attempted.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.