Guillermo's suggestion to use a sparse-file based portage has given me an idea. I can simply create a second VMWare disk that will, if present, be loaded into the system, otherwise /usr/portage will be empty. People who want to update or add software to the system can download the portage disks.

That will also mean that I can work on the image without all the disk-churn of deleting and re-installing the portage tree every time I want to add or update something for the developers. I'll probably put both the working /var directories and /usr/portage on the secondary disk. There will still be growth due to the replacements, but it won't be as insane as with all the work happening on the main image.

I know, I know, I should probably be working on the Fedora Core version, but I don't think I'll finish that before tomorrow and I want to have time to burn 15 or so developer's images tonight (I'm off on a client site-visit all day tomorrow before the presentation).

The cp based reduction of the vmware image worked well, though I had to re-install grub on the new partition using the Gentoo handbook approach. I spent a while looking into the use of aufs, it looks promising, but I'm not yet confident enough with it's usage to try to deploy it to users.