I apparently had a misconception that during a REORG SHRLEVEL CHANGE, the table is only unavailable during the last LOG iteration and SWITCH phase.
It appears that it is unavailable during UTILTERM also. The master log was showing some -904 resource unavailable, however, the display database showed the tablespace in UTRW (I don't know if UTRW status was for the entire UTILTERM phase or if the -904's occur only during a portion of the phase).
I have several more tablespaces to convert and my concern is that some of them will have a very long UTILTERM phase because of several indexes with small piece sizes (for pre-Db2 9 REORG Performance) and it takes what I think is unreasonably long to delete the old datasets (several seconds each), and it deletes serially.
Is there a way to speed up the delete? I've asked the Systems Programmers and they don't think several seconds for a single delete is abnormal. I am considering doing an ALTER on the indexes to increase the PIECESIZE to reduce the number of datasets involved.