TEMP tablespace getting full while inserting a CLOB in Trigger

Question

We have a Oracle 10g (10.2.0.4.0) DB on a Solaris 9 box which also runs our J2EE web-service application on Weblogic 8sp6 server.
We get around 220K web-service requests from upstream callers daily to insert data in the main table, say TABLE1, which has daily partitions on a date column. This table has around 21 columns out of which 1 is a CLOB column.
Now this table has an AFTER INSERT trigger which calls a package procedure to insert the same record into another table, say TABLE2.
From Java application insert statement in executed in below format using a weblogic jdbc connection pool :
INSERT INTO TABLE1(COLUMN1, COLUMN2, ........., CLOB_COLUMN,........, COLUMN21) VALUES (:1, :2, :3, :4, :5, :6, :7, :8, :9, :10, :11, :12, :13, :14, :15, :16, :17, :18, :19, :20);
Clob object is prepared in application using ojdbc14.jar.
We are observing a strange issue here. The TEMP tablespace utilization keeps on growing as more and more inserts are executed by application and after ~125K inserts the TEMP tablespace gets full and we start getting ORA-01652 error.
On further analysis we could see that there are only 7-10 session being maintained but as more and more inserts happen TEMP tablespace utilization goes on increasing for each of these sessions.
When we tried with inserting just few records and then watching the session details in v$session_wait then we could see that it is in INACTIVE state and waiting for the event ‘SQL*Net message from client’. This does not seem correct as the session has successfully inserted the data and committed the transaction and we can see the data in the tables as well.
The confusing thing here is when we modify the trigger to pass blank string('' ) instead of the CLOB column to TABLE2 then this issue does not occur. All 200K records are inserted properly and TEMP tablespace utilization also keep always below 1%.
Can you please help us in solving this issue. Is this related to any oracle issue?
Inside the package we have tried using DBMS_COPY statement to copy the CLOB column after insert but still same result.
Code for reference:
Trigger:
=====================================
CREATE OR REPLACE TRIGGER trg
AFTER INSERT OR UPDATE
ON TABLE1
REFERENCING NEW AS NEW OLD AS OLD
FOR EACH ROW
BEGIN
IF (:NEW.date_col > SYSDATE - 2)
THEN
IF (:NEW.cat IN (1001, 1002))
THEN
pkg.process_change
     (:NEW.COLUMN1,
          :NEW.COLUMN2,
          :NEW.CLOB_COLUMN,
FLAG
END IF;
END IF;
END;
=====================================
Package:
=====================================
procedure PKG.Process_change(
p_COLUMN1 number,
p_COLUMN2 varchar2,
p_CLOB_COLUMN clob,
flag boolean
) is
v_watermark pls_integer;
v_type varchar2(1);
begin
if (flag) then
v_type := 'U';
else
v_type := 'I';
end if;
select t_seq.nextval into v_watermark from dual;
insert into TABLE2(
COLUMN1 number,
COLUMN2 varchar2,
CLOB_COLUMN clob,
watermark,
dml_type
)values (
p_COLUMN1 number,
p_COLUMN2 varchar2,
p_CLOB_COLUMN clob,
v_watermark,
v_dml_type
end;
=====================================

Answer 1

My first thought on reading your post is that not only are you using a database version that is now so old it is in extended support and even then not even the most recent patchset for it.
The first thing I would do is move to 11gR2 and if you can't do that at least get to 10.2.0.5 and apply CLOB relevant patches as well.
Same goes for your operating system. Solaris 9 is ancient: So move to 10 which has vastly improved memory management.
To help you further it would be really valuable to know the table layout. For example is this a heap table or an IOT? Is it partitioned? Is this RAC? What size are the CLOBs? Are they stored in-line? Chunk size? etc.
This page should start you down the right road:
http://docs.oracle.com/cd/B19306_01/appdev.102/b14249/adlob_tables.htm#sthref204
But I am also wondering why you would use a trigger to, as you say, "insert the same record into another table." This description is a poster child for "bad design."