negative elapsed times in 10046 trace file for single block reads

10g R1 std ed 32 bit (10.1.0.4 with cpuoct2006 applied). w2k3 R2 sp1 32 bit
CPUs: a pair of dual core AMD Opterons
datafile storage is on a NetApp Filer attached by a pair of non-TOE enabled onboard gigabit ethernet adapters using an MS iSCSI driver.

I'm seeing negative values for elapsed time in a 10046 trace file (lots of them, actually):

WAIT #27: nam='db file sequential read' ela= 6652 p1=24 p2=52089 p3=1
WAIT #27: nam='db file sequential read' ela= -365131103 p1=25 p2=58558 p3=1
WAIT #27: nam='db file sequential read' ela= 15075 p1=25 p2=58560 p3=1

A quick search of metalink returned only the reference doc *Note:39817.1 **Interpreting Raw SQL_TRACE and DBMS_SUPPORT.START_TRACE output*

Has anyone else run across this before?

thanks,

Paul

--
http://www.freelists.org/webpage/oracle-l