On FreeBSD, where ports allows you to build from source, kde will go into /usr/local. In your environment, /usr/local would be mounted from you new shiny 1TB drive.
If filesystem is on (non-redundant) disk(s) that break, then some/all of that filesystem is missing. Say, filesystem spans two disks and one breaks. Which parts are in the intact disk? Can you (trivially) read anything from it?
Compare that to having separate filesystem in each disk. The intact disk will still have consistent and complete filesystem with whatever files are in it. Only the content of the broken disk is lost.
One does not need symlink, if one mounts to correct point. For example, we know that we will have a large webpage. That data would be under /var/www.
We could just mount a second filesystem (i.e. disk) somewhere and reconfigure the HTTP-server to use that new path instead of /var/www.
We could mount somewhere and replace /var/www with a symlink to new path.
In these HTTP-server config and or AppArmor/SELinux has to be tweaked.
Or, mount new filesystem directly as /var/www.
I actually would mount elsewhere, create subdir for www content and then bind-mount that subdirectory as /var/www.
I would be very careful about installing OS components (like KDE) on a portable storage device unless you can insure that that "portable storage device" is always properly connected when booting the computer.