This is caused by the client (or agent) not requesting enough pre-allocated
space in the diskpool for the data that it "actually" needs to send.  We ran
into this on both Tandem clients and SQL agents.  On the SQL there is an
"estimated change percentage" parameter that you can use and it will
pre-allocate that amount of space in the diskpool.  On the Tandem client you
set "objectsize" and it preallocates the space accordingly.  If you are
using an "agent" I would look for something like that.  If it is a B/A
client then you should look to see if compression is turned on.  This could
possibly cause the data that is sent to the server to be larger that what it
was originally calculated at.