I too have seen this behavior after migrating to 3.1.2.1.  The
expiration,
which used to run successfully in about one hour, now takes days
utilizing
as much cpu as it can get.  The object counts in the status message
stop
incrementing after about an hour.  What is it doing between hour 1 and
hour
100?  Is this a bug?  Can I safely cancel the process after the counts
stop
incrementing and not lose anything from expiration?   IBM --- can you
comment please?