Description of problem:
This problem is found when debugging vdsm.
If a libvirtd client register its event handleImp to libvirt, then connect to libvirtd, then call os.setgroups, libvirtd connection will be closed in 30 secs after os.setgroups called, seems a connection keepalive scheme problem, but connection is available in this 30 secs.   
 

Version-Release number of selected component (if applicable):
libvirt-0.9.11-1.fc16.x86_64

How reproducible:
see demo script in attachment(using sudo)
(1)start a event loop and register the handleImp to libvirt
(2)open a connection to libvirt
(3)call os.setgroups([])
(4)call libvirt fuction
(5)sleep 40 secs
(6)call libvirt function
  
Actual results:
after step (3) we got SIGCHILD which interrupt poll
step (4) will gain right result
step (6) will result in broken pipe:
libvirtError: Cannot write data: Broken pipe

libvirtd error log:
2012-05-23 12:17:42.612+0000: 2773: warning : virKeepAliveTimer:182 : No response from client 0x132d500 after 5 keepalive messages in 30 seconds
2012-05-23 12:17:42.612+0000: 2773: debug : virConnectClose:1496 : conn=0x7fea6c0089e0
2012-05-23 12:17:42.613+0000: 2773: debug : qemuDriverCloseCallbackRunAll:661 : conn=0x7fea6c0089e0

Expected results:
connection should not be closed, keepalive scheme should active normaly after os.setgroups()

Additional info: