The findCenterPositionsOneTrace routine identifies one trace per call. It starts each time by stepping up through rows looking for the start of an "aperture": a wide-enough and bright-enough span of columns, then tracing the full aperture, marking it in the image, then returning. The logic to continue searching after hitting an old or a failed aperture match was not able to find multiple apertures which started at a given row.

Given the tricky book-keeping and logic, I now pass the next unexamined pixel on the first matched row back to the caller, which then passes that back in to findCenterPositionsOneTrace to initialize the next aperture search.