Plan stability

I have two execution plans for the same query from two different machines.
Both are running on Tru64 v5.1, Oracle 8.1.6.3
SORT_AREA_SIZE for plan1 is 47MB, plan2 is 12MB

*************************************************

1)

SELECT STATEMENT   Cost = 63297

  NESTED LOOPS

    NESTED LOOPS

      HASH JOIN

        TABLE ACCESS FULL TABLE4

        HASH JOIN

          TABLE ACCESS FULL TABLE3

          TABLE ACCESS FULL TABLE1

      TABLE ACCESS BY INDEX ROWID TABLE5

        INDEX UNIQUE SCAN SYS_C001602

    TABLE ACCESS BY INDEX ROWID TABLE2

      INDEX UNIQUE SCAN SYS_C001696

************************************************

2)

SELECT STATEMENT   Cost = 45571

  NESTED LOOPS

    NESTED LOOPS

      MERGE JOIN

        SORT JOIN

          MERGE JOIN

            SORT JOIN

              TABLE ACCESS FULL TABLE3

            SORT JOIN

              TABLE ACCESS FULL TABLE4

        SORT JOIN

          TABLE ACCESS FULL TABLE1

      TABLE ACCESS BY INDEX ROWID TABLE5

        INDEX UNIQUE SCAN SYS_C002545

    TABLE ACCESS BY INDEX ROWID TABLE2

      INDEX UNIQUE SCAN SYS_C002617

*************************************************

There are approximately 200000 more records in all the tables for plan2 than for plan1 tables.
1) Why is plan2 sorting?
2) How does the init parameter SORT_AREA_SIZE effect the plan stability?

Are you sure that sort_area_size is the only factor? What about hash_area_size ,hash_join_enabled ?
There some other "hash" setting in init.ora which being used by optimizer to decide what to use: sort or hash join.
My point is that make sure that all other init.ora parameters are exactly the same except sort_area_size.