Wallet INR taps were priced at a provisional 0.0180 AUD/INR; CommBank charged
~0.0149, so every rupee twin sat ~20% outside the picker's +-5% band. The
auto-matcher pairs on the native figure but refuses ambiguity (INR 112 twice in
three days, INR 175 three times), which left those rows offered nothing at all.
getPendingReconciliations filtered neither the manual side nor the candidate
statement side on superseded_by_id. After the 2026-10-03 wallet cleanup the
queue still showed duplicate half 7037 (superseded under 7038) beside its
twin, with the same Digidirect statement line offered to both -- accepting it
on the hidden row would have spent the line and stranded the real one. The
candidate side could likewise offer one of the 31 superseded ANZ rows.
findAutoMatches() always had both filters. First tests for this query;
both fail against the old SQL.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>