TAF: instead of waiting, ORA-00028: your session has been killed

Question

I'm testing the ability of our application to work with a RAC cluster.
When I stop a node with the command:
srvctl stop service -d DB_UNIQUE_NAME -s SERVICE -f, I expect that the driver will try to reconnect transparently without notifying the application.
In some circumstances, the application may receive errors in the range ORA-25400 - ORA-25425, but the reconnect should be automatic anyway.
This is the case, mostly.
But the last time when I stopped the node, 9 of 10 threads of our app began to wait for the driver's reply (as expected). And one thread instantly received the error: ORA-00028: your session has been killed
The subsequent attempts to use this connection failed with: ORA-01012: not logged on
tnsnames.ora:
PPB_RAP.EXASTAG.PIOSCS =
  (DESCRIPTION =
    (CONNECT_TIMEOUT=4)(TRANSPORT_CONNECT_TIMEOUT=3)
    (ENABLE=BROKEN)
    (ADDRESS_LIST =
      (ADDRESS = (PROTOCOL = TCP)(HOST = sdm2-scan.it.bwns.ch)(PORT = 1521))
      (ADDRESS = (PROTOCOL = TCP)(HOST = sdm1-scan.it.bwns.ch)(PORT = 1521))
    (LOAD_BALANCE = yes)
    (CONNECT_DATA =
      (SERVER = DEDICATED)
      (SERVICE_NAME = PPB_RAP.EXASTAG.PIOSCS)
      (FAILOVER_MODE =
        (TYPE = SELECT)
        (METHOD = BASIC)
        (RETRIES = 180)
        (DELAY = 5)
  )Client: 11.2.0.1.0 32bit
Server:
Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options

Answer 1

940607 wrote:
What does it mean?http://www.lmgtfy.com/?q=oracle+TAF+DML