Failure to start a jail successfully leads to lingering epair interface on host while the jail (and its epair interface) disappear. Destroying the lingering epair interface on the host leads to a kernel panic second to minutes later.
Can you send a jail.conf and commands that you use for replication? I just tried this on my machine and it worked fine.
Created attachment 243902 [details] /etc/jail.conf bastille jail
Created attachment 243903 [details] logs0/scripts/created
Created attachment 243904 [details] create-bridge.sh
Created attachment 243905 [details] bridge-jail-epair.sh
Created attachment 243906 [details] create-jail-epair.sh
1. Start a jail 2. Using exec.created/exec.poststart, successfully move the epair interface into the jail via ifconfig vnet. 3. Run another script via exec.created/exec.poststart, but have it fail, causing the jail's creation to fail and exit. 4. Observe that the first epair interface still exists on the host via ifconfig. 5. Some seconds to minutes following this, the kernel will panic and reboot the system.
(In reply to Antranig Vartanian from comment #1) ping