I let imapsync run overnight to migrate my ver big IMAP message store to Zimbra.
This morning I found the forcly stopped imapsync with a "Broken pipe" error message. I could not contact Zimbra using an IMAP clien or the web client.
On my Zimbra server /etc/init.d/zimbra status say that "mailbox" is stopped.
This relevant last lines in zimbra log say
Zimbra should handle such things more gracefully.
2005-09-11 03:02:56,923 FATAL [ImapServer-2]  system - Fatal error occurred while handling connection
java.langOutOfMemoryError: Java heap space
Thanks for the report. Can you provide some information about your hardware? How much memory? CPU size/speed. At this time we aggressively exit on out of memory to prevent any data corruption. This is a known issue and are working on change it.
The bare metal is a cheap PC with a Sempron 2200 (1.4 GHz), 768 MB RAM, an average 60 GB HDD running Windows XP SP2. The GPU steals 64 MB of the RAM = 704 MB available.
I'm controlling this PC remotely from my PowerMac G4 Cube using Microsoft Remote Desktop. I've a 100MB/s switched LAN.
The VMware Workstation 5 box has 512 MB RAM and a virtual 8GB HDD* assigned. I installed Fedora Core 4.
With Zimbra running in that box and after migrating the first 800 MB of my IMAP message store the file system was 81% full**. After deleting files from the redo log archive the fs was 70% full.
* I added another virtual 8 GB HDD after the crash. Hence the delay of my answer to your question.
** Estimated values from (brain) memory.
Update here. We are going to tweak the way we default some system params on installs will less than 1GB MEM. This should help a bit.