I'm experimenting with TreeCache and Hibernate. I have things running (it took me a few days to work up a fairly simple example), but I think I've encountered a stumbling block:

I'm running on Windows XP. If I configure TreeCache with CacheMode set to INVALIDATION_SYNC, and the loopback attribute on the UDP element of the ClusterConfig is set to true (as is documented to be required on Windows), then I do get cache misses for updated objects in other VMs (Good, as expected), but I /also/ get cache misses for updated objects in the VM that originated the update! This is bad, as it effectively makes the cache mostly useless.

I know that turning IP Multicast loopback off in Windows isn't supposed to be possible, but is there any way to work around this limitation of the OS? All of our development boxes are Windows, and it would be /really/ bad to have behavior on development not consistent with behavior we see once deployed.