The problem is that once the tape is deleted from the pool, it returns to
scratch.  The next scratch mount calls for the same tape.  This was addressed
by APAR  IC17476, and the fix is that  all the scratch tapes are now cycled
through instead of selecting the same bad one over and over.  So the second
time, it should select another scratch IF there is one.  If there isn't, it
will select the bad one again.  This was intentional, as otherwise the tape
could forever be left in the library as scratch but unusable and taking up
space. This way, the bad tape would intermittently show up with I/O errors thus
drawing attention to the need to remove it.

Craig Bell