----- Original Message -----
I guess you could add some calls to trace the output of the various clock functions around there, and see how the monotonic clock drifts over time?
That is a bit out of my knowledge/comfort level. However, after further testing, I've found a later kernel/Glibc solves the problem. When running the CentOS 6 container on the following system, I have the errors present: Kernel: Linux virt3 2.6.18-6-pve #1 SMP Mon May 9 13:27:11 CEST 2011 x86_64 GNU/Linux Glibc: 2.7-18lenny6 However, after migrating the CentOS 6 container to a more updated system, the errors go away: Kernel: Linux virt1 2.6.32-4-pve #1 SMP Mon May 9 12:59:57 CEST 2011 x86_64 GNU/Linux Glibc: 2.7-18lenny7 My (somewhat educated) guess is the changes to the CentOS 6 kernel in regards to timing are also reflected somehow in the software. The kernel being used is not from CentOS 6 (but rather Debian, patched with OpenVZ), and the later kernels support or handle whatever it is the CentOS 6 libs/ABI/whatever are looking for. Does that make sense? --Tim