A possible cause of this abend is that the OPSOSF server has a WLM Class higher than the OPSMAIN address space WLM Class.
OPSMAIN task priority should be comparable to JES. At least it should be set to SRMGOOD. If it runs at the lower priority, there is a small window where OPSMAIN task can be suspended after starting a server and when this occurs the ACBINRTN field is not filled in for the server subsystem dataset.
The problem is a timing problem introduced by local settings for the OPSMAIN address space in WLM. Since it is a low priority when it issues the ASCRE to create the OSF server, it is suspended by WLM, and the OSF server attempts to open its subsystem datasets, but OPSMAIN did not fill in the ASCB address of the server because of the suspension. So when the OPSS10 exit is driven, it cannot identify the address space trying to open the subsystem dataset.
The documentation reports that OPSMAIN should run at the same priority as JES2, and normally OPSMAIN should be set to
SERVICE CLASS = SYSSTC
WORKLOAD = SYSTEM
INTERNAL CLASS= $SRMGOOD
Anything lower could result in the abend described before.
The OPSOSF/TSF Servers can remain lower, unless they will start having performances issues.
Workload Class is the criteria that defines which group of tasks get prioritization in the system. The Dispatch Priority defines prioritization within the class.