We have a WebSphere cluster with two nodes running with a dual message busses configured in a high capacity configuration. Doing some tests with our application we see the application connection status getting in trouble and it appears that trying to shut down the server takes anywhere from 40 - 60+ seconds where a server without any blocking connections established takes 6 seconds to shut down.

Some sample logging statements show that the delay is actually prior to the destruction of the servlet, which seems to indicate that the delay isn't anywhere in client code.

[11/26/09 12:00:50:265 MST] 0000002f AdminHelper A ADMN1021I: An attempt is made to stop the lntest9Member server.
[11/26/09 12:00:50:341 MST] 00000012 TCPChannel I TCPC0002I: TCP Channel TCP_2 has stopped listening on host * (IPv6) port 9082.
[11/26/09 12:00:50:344 MST] 00000012 TCPChannel I TCPC0002I: TCP Channel TCP_4 has stopped listening on host * (IPv6) port 9445.
[11/26/09 12:01:39:022 MST] 00000012 ApplicationMg A WSVR0217I: Stopping application: failoverApp_noeps_war
[11/26/09 12:01:39:060 MST] 00000012 ServletWrappe I SRVE0253I: [failoverApp_noeps_war] [/failoverApp] [Blocking Servlet]: Destroy successful.