Resolving The Problem

When any task OPENS a PS, VB or VBA dataset, such as the TWSz EQQMLOG for OUTPUT with DISP=SHR, that task starts writing to the dataset at the very beginning, overwriting whatever was in it before. As it continues writing, it may over-write the EOF (End Of File) marker, thus leaving the dataset with no valid EOF indicator. This marker will be re-written when the dataset is CLOSED by the owning task, so the situation is transitory, and completely normal.
.
However, if some other task READS the dataset while there is no EOF, the READING task may read past the valid data in the dataset into space that is allocated, but has not yet been written to. This is a normal exception condition, but is reported by z/OS as an I/O error, with messages such as IEC036I 002-04.
.
Similarly, if the dataset is BROWSED via ISPF while it has no valid EOF indicator, ISPF will report the same I/O error by displaying the following line after the last valid (readable) record:
.
***** * * * I/O error detected, I/O terminated * * * *****
.
This too is completely normal, correct processing. and the indicated I/O error can be ignored.
.
A related symptom is when you scan down the EQQMLOG and find yourself looking at "old" messages... written by TWSz several days ago. In this case, you have read past the data written by TWSz since it was last started and are looking at valid data from a previous run. The cause is exactly the same, although the symptom is different.