lld_process_lost_objects() function uses 'per field' update approach. It creates multiple id vectors for different field values types and then executes single field update per each vector, for example:

update item_discovery set ts_delete=1718646734 where itemid=47167;
update item_discovery set status=1 where itemid=47167;
update item_discovery set disable_source=1 where itemid=47167;

This also makes harder to update relevant fields - ids might need to be removed from some vectors and added to others.

LLD mostly uses 'per object' approach - all changes are gathered for each object, flagging the updated fields and then single update query for each object is generated, for example:

update item_discovery set ts_delete=1718646734,status=1,disable_source=1 where itemid=47167;