Here's a real gotcha. IBM Levels 1 and 2 are aware of this problem, but cannot supply a fix. Maybe someone out there has a better workaround.
If you configure a device, say on port 2, address 2 of controller 1, and you attempt to make the new device (in this case a 3196 workstation) available, you may get the message "(Workstation ID) cannot be made available."
If you display the error messages in the job log, the primary message is CPD2896 ("Device [Workstation ID] vary on stopped"). The cause is listed as "device (Workstation ID) is not operational," and the recovery suggestion is to check cabling and switches for possible hardware problems.
Now the real fun begins! Even if you delete the new device description, you will be unable to configure any devices on that particular port until you IPL the system again! It seems that the good old system arbiter (QSYSARB) has an object lock on the deleted device description and will not let go, no matter what you do, short of an IPL.
Our local SE explained that performing another IPL is what he does when this situation arises, and Levels 1 and 2 have no fix in their database except an IPL. They suggested that shutting down QINTER and restarting it might help, but this did not work for us.
We're on V2R1M1 and have no Authorized Program Analysis Report (APAR), which generally precedes a PTF. As of November 1, Level 1 had no fix for this QSYSARB glitch.