This is a partitioned table with 55 millions of row. The "repair" is two passes process. First pass to collect out of rows and second pass to repair those rows. It helps to minimize the target table locking time. compare client was waiting for compare marker, eventually, the compare client gave up the second rc (read consistent) point because it has to search all rows.
Large table takes to long to scan and compare/repair fails since it can get to the second read consistency marker.
It is highly recommended to compare partition by partition:
sp_ctrl (alvclbu02:8444)> compare
Usage: compare table {source_owner}.{source_table}[@partition name] to
{target_owner}.{target_table}[@partition name] target {target_host}@o.{target_sid}
