使用Merge优化改写一个update语句

业务需求需要将错误数据进行整理，主要是针对于length(l.brchno) >= 4这部分数据，以下是一个UPDATE进行更新的操作。
源表有28万的数据量，符合条件的有54327条数据。UPDATE执行时间大概是一个半小时，使用Merge进行优化后执行时间是27秒
explain plan for
update sif_member_load l
   set l.brchno =
       (select brchno
          from (select distinct d.brchno, d.brchna
                  from sif_member_load d
                 where d.brchna in (select brchna
                                      from sif_member_load ld
                                     where ld.brchno like '%寿险%'
                                        or ld.brchna like '%寿险%'
                                    )
                   and length(d.brchno) < 4
                   and d.brchno <> '8619') k
         where l.brchna = k.brchna)
 where length(l.brchno) >= 4
   and l.brchno not like '%综拓%'
   and exists (select distinct d.brchno 
                  from sif_member_load d
                 where d.brchna in (select brchna
                                      from sif_member_load ld
                                     where ld.brchno like '%寿险%'
                                        or ld.brchna like '%寿险%'
                                    /* where l.brchno = '寿险续期'
                                     group by brchna
                                    having count(*) > 1*/
                                    )
                   and length(d.brchno) < 4
                   and d.brchno <> '8619');----加exists原因是找到符合条件的数据

select * from  table (dbms_xplan.display);
 执行计划
SQL> select * from  table (dbms_xplan.display);

PLAN_TABLE_OUTPUT
-----------------------------------------------------------------------------------------------------------------
Plan hash value: 2740296014

-------------------------------------------------------------------------------------------------
| Id  | Operation