I know you were mentioning lxc earlier, just thought I'd chip in on a little observation from using it with my Jessie/Hildon GTK3 side project via systemd (can't remember exact arguements but it uses systemd-nspawn).

Firstly, accessing is so much easier than messing around with chroot. /sbox/login does a little more but I wanted something actively maintained. shroot and a few others didn't appear to be seeing much development when I started it. Networking wise I did nothing, it just worked.

I love the fact you can "boot" the instance, makes it easier to test packages starting correctly on boot BUT it doesn't fully emulate all device nodes. One package wouldn't build unless adding extra options to bind /dev/input. Another, possibly statefs, failed to start, causing "boot" to take forever or in some cases cause the container to constantly reboot. There are some options to enable more capabilities/CAPS but I don't fully understand all of them at the moment. As a result I no longer use this mode very much.

Currently I use the chroot equivalent to build and then have a script to mount the directory as a nfs share so I can boot a qemu vm using the same root folder without having to constantly build fs images. networking here needed more configuration host side.