After the Bionic 5.0 HWE kernel was released I recreated my Disco container to use shiftfs. After running fine for several days, this morning I found that processes running as part of a manual apt update (like gpgv) were hanging in D state. I was able to kill them with ^C. Stopping the container took a long time as well. This is with LXD 3.15 (11437).
So I deleted the container and created it without shiftfs and will watch for the problem to recur. I had no problems running this Disco container with the previous Bionic 4.18 HWE kernel. I use a “real” ZFS for container storage and a “real” bridge with its own network interface. The only unusual part of the configuration is that the container is on a WireGuard VPN using the kernel module from the host.