Intermittent ora-1017; poor workaround lsnrctl stop/start

Question

Running Oracle 11.2.0.3 on a redhat Linux x86_64 installation.
I've mapped the servicename to the desired sid in the listener.ora to avoid continually remapping the config information in the source code. This allows the URL to remain consistent for all developers, "jdbc:oracle:thin@//host:port/service" and during predeployment testing it is changed in the listener.ora file to reflect the "production" deployment for final testing.
This has been working great until recently when we began to receive intermittent ora-1017 during connection. If I stop and start the listener, lsnrctl stop:lsnrctl start, I can connect successfully for two sessions and then the ora-1017 reappears.
os authentication is in use(until ??? & I know it's bad) with remote_os_authent=true os_authent_prefix=''. User accounts are the same in both instances, "scott identified externally" for instance
The client is Windows XP and the connection is made using hibernate 3.26 (empty pwd/uid in hibernate.cfg.xml)
listener.ora:(example)
sid_list_listener =
(sid_list =
(sid_desc =
(global_dbname = productionServiceName)
(ORACLE_HOME = /the/oracle/home)
(sid_name = developmentName)
(sid_desc =
(global_dbname = developmentName)
(ORACLE_HOME = /the/oracle/home)
(sid_name = developmentName)
two connections and ora-1017 consistently.
I've noticed that the lsnrctl status reports an increased number of handlers as soon as this error occurs for the "productionServiceName" but I'm unsure whether theere's a connection to the error
host > lsnrctl status
Service "productionServiceName" has 1 instance(s)
Instance "developmentName"status UNKNOWN, has 1 handler(s) for this service
ORA-1017...
host > lsnrctl status
Service "productionServiceName" has 2 instance(s)
Instance "developmentName"status UNKNOWN, has 1 handler(s) for this service
Instance "productionServiceName", status READY, has 2 handler(s) for this service
Thanks for any help

Answer 1

Very insightful response, thanks :|. I asked the same question of our administrator and she insists she changed nothing.
However, I found that dynamic service registration via PMON/service_names init parameter and what appears to be "load-balancing" is causing the 1017 error regardless of what I specify in the listener.ora. According to the oracle docs (link here) the dynamic service registration takes precedence over the static sid list entries in the listener.ora. The invalid username/password error was occurring because the "production instance" had remote_os_authent set to false, so as the uncontrollable load balance switch occurred the username of course is invalid.
I asked our administrator to set the remote_os_authent to true on the "production" instance and now i receive a different failure - intermittently - due to the absence of a collection object in that instance accessed during application launch.
Now to determine when the switch is occurring and control it OR find an alternative to allow constant jdbc configuration with a dynamically configured back-end...as well as to determine what allowed this to work in the first place?
My configuration was based on the statement in the listener.ora configuration documentation that says:
"Use the parameter GLOBAL_DBNAME to identify the database service.
While processing a client connection request, the listener tries to match the value of this parameter with the value of the SERVICE_NAME parameter in the client connect descriptor. If the client connect descriptor uses the SID parameter, then the listener does not attempt to map the values" It does go on to state this is mainly for Oracle8 , 7 but does not indicate issues with newer releases nor that it may be OBE.
****** SOLUTION I FOUND *********
is to use the new 11G parameter DEFAULT_SERVICE_<listener>=<serviceName> configuration in the listener.ora and to NOT specify a service in my jdbc connect string, jdbc:oracle:thin:@//host:port. This provides the functionality I need cleaner than the initial config -
Thanks to all who helped