We weekly IPL S0W1 and then S0W2.  We occasionally need to IPL S0W1 without IPLing S0W2 during the week.  We take down S0W1 and take it out of the plex and S0W2 continues to run fine.  When we IPL S0W1 and it starts to join the plex (as indicated while still in the NIP console), S0W2 immediately loses TCPIP connectivity.  We can logon to S0W2 via a local terminal during this time with no problem. We have to recycle VTAM, TCPIP, TN3270, etc, so we usually just IPL it.   
Is there some dependency of which I am unaware?  This does not happen on a "real" sysplex.                                                   
.       

Sherril, I realize you appended to this forum after a response to the PMR you opened for this. Since it is not clear whether this is a z/OS configuration problem or a zPDT issue, Mike Hockings and I will pursue this through the PMR you opened for it. We will be asking some questions in the PMR, and additionally requesting some system logs and display commands. After we get some resolution, I will append the results to this forum for others who may encounter this.

Just as a follow-up here is some information for anyone else that may hit this situation:

The key is that the two z/OS systems in the sysplex are sharing an OSA adapter in the devmap.  The port names in the VTAM TRLEs on both systems must be the same (such as PORTA or PORTB) in this case.  The first system to open the port, which would be the first z/OS system up (S0W1 in this case), then controls that port name even if that first system is later stopped.  In other words, the OSA is working as designed, even though connectivity to the second z/OS (S0W2) is lost if the first z/OS (S0W1) is IPL'd.

The issue may be bypassed by defining independent OSAs for each system but this has not be tested to confirm this hypothesis.

As a work-around the following commands on S0W2 should help restore connectivity : 

V TCPIP,,START,PORTA
V TCPIP,,START,PORTB

The bottom line is that the RD&T environment is working correctly.