Multicast longevity problems

When running multicast we have noticed some stability issues - multicast works fine for a few hours but the just stops working/passing traffic. Unicast is OK (we can ping all hosts from each other but multicast is a no-go). Rebooting the VPNCubed (Free Edition) instance usually resolves the issue right away. Is there some separate process for multicast on there (some routing or RP perhaps?) that we can troubleshoot?

Our hosts are all CentOS 5.3 x86_64 - the same software/OS/build works fine off of EC2 where multicast is natively supported.

We had thought perhaps there was a DF issue so we lowered MTUs for the tunnel interface on the hosts but this didn't resolve the issue either.