Please stop hardcoding 127.0.0.1 in all parts of zimbra. Really makes it quite impossible to run Zimbra in linux-vservers without a myriad of changes throughout.
So far, it's still possible with lots of manual hacking to get it working (see the wiki page on linux vservers i've largely kept up to date), and using some other RFC1918 IP for localhost in /etc/hosts, but now the install script in 5.0.2GA is requiring 127.0.0.1 to be localhost at install time.
It's really too bad that localhost has been assumed to be statically set to 127.0.0.1 since the dawn of time, that's not always the case.
Linux vservers share a network stack (unlike the extremely heavyweight Xen and VMWare virtualizations), so 127.0.0.1 as a specific static IP belongs to the vserver host and cant be used (easily) by the vservers.
Wether vserver is a viable virtualization technology or not is outside the scope of this discussion (except to note that vservers are quite popular and viable, and far more efficient than vmware or xen), and considering what small changes in configuration are required so far to keep it working JUST FINE in vservers, it'd be a shame to see an errant 127.0.0.1 somewhere cause it to STOP working for the wont of a few bytes.
Virtualization is a hot topic now, and with power costs soaring along side CPU power and #s of cores and the plummeting price of ram, I dont think it's going to go away - please dont lock off one very nice avenue of virtualization for Zimbra in the linux world.