Probes staged here are NOT yet in validation/ because they have no tv_trades.csv reference yet (capture pending). They are excluded from verify_corpus.py --all (which scans basic/community/validation) so they do not affect the headline parity count. Each carries strategy.pine, generated.cpp, and an engine_trades.csv (current engine output on the 15m chart feed derived from data/ohlcv_ETH-USDT-USDT_1m.csv). Once a TV export is captured into a probe's tv_trades.csv, move the dir into validation/ to activate it.

Designed 2026-06-28 to test the hypothesis that process_orders_on_close=true drops a from_entry strategy.exit bracket issued while flat (alongside a pending limit entry that fills on a later bar) — the shape of community initial-balance / breakout strategies, and a suspected cause of exit divergences observed in engine parity testing.

RESULT: hypothesis REJECTED. On the reference feed the engine output is IDENTICAL for process_orders_on_close true vs false (179 trades each; 12 bracket fills [11 stop + 1 TP], 167 opposite-cross). The from_entry exit survives the placement→fill gap and fires at its level under both modes — no on_close-specific drop. (An earlier 5-bar synthetic suggested a drop; that was a contrived-feed artifact, disproven by these real-feed probes.)

Kept as regression coverage for the process_orders_on_close × limit-entry × exit-issued-while-flat pattern, pending TV capture to confirm engine↔TV parity.