We have an ADF application, which hooks into Portal and SOA running on Weblogic 10.3.6 in a Fusion Middleware environment.
A few weeks ago after pushing a new release of the code to production, users encountered an ORA-01722 error in some screens of the application. We captured the SQL and could run it via SQL*Plus and Sql Developer with no errors.
We opened an SR with Oracle and they offered some suggestions and in the end recommended a patch to fix a shared cursor bug. We haven't applied that patch yet because we are on the latest PSU for 184.108.40.206 and that patch isn't available for it. So we've issued a patch request with Oracle.
The weird problem is, that when my co-worker starts the WebLogic Managed Server, the users do not encounter the ORA-01722 error. The only difference is that I am ssh'ing from a Mac and he is on windows. We ssh into the same generic account and execute the same steps. We've also confirmed that I can restart the server a dozen times with the error and my co-worker has never had the problem. We've verified this a few times.
Today we had some more maintenance and I tried restarting the server again. Users encountered the same ORA-01722 error. I restarted the server again but this time from my windows computer using SecureCRT. Error disappears.
The following is a diff between a mac and windows ssh session. There isn't a huge difference.