I've been running GPUGRID for a while with no issue but recently I am seeing all the KASHIF_HIVPR units erroring out. They mostly error out well into the WU (anywhere from 1,000 seconds to 15,000 seconds in), not at the outset. TONI workunits (KKi4, CAPBIND, and MERGMPO2) are generally still working fine. I'm not too worried about this, since it's still generally busy with the TONI WUs, but any idea what is going on?

Linux x86_64 kernel 2.6.35.7 (Fedora 13)
BOINC 6.10.58 (x86_64)
NVIDIA Linux x86_64 260.1912
GeForce 9600 GSO (512 MB)
2GB system RAM (<50% used)

This machine does almost nothing else besides GPUGRID.

Machine profile: http://www.gpugrid.net/show_host_detail.php?hostid=75166
Result list: http://www.gpugrid.net/results.php?hostid=75166


Example WUs with these errors:

http://www.gpugrid.net/result.php?resultid=3143972
http://www.gpugrid.net/result.php?resultid=3133613
http://www.gpugrid.net/result.php?resultid=3128437
http://www.gpugrid.net/result.php?resultid=3128120
http://www.gpugrid.net/result.php?resultid=3117902
http://www.gpugrid.net/result.php?resultid=3113802
http://www.gpugrid.net/result.php?resultid=3112420