In v4.2 we created cursors as needed, so we would create a cursor to scan an entire index or collection table.

In v4.4, we now create all cursors up front and reuse them errors are found to run the second phase (same as the first phase except saving more information). So rather than holding a snapshot for the duration of a single table scan, we could end up holding a snapshot open for 3x (say one collection scan and two index scans), or 6x (if the corruption was found and the second phase runs all the validation again).