Indeed, it does not seem to be tied to a specific upstream resolver.
I won't rule out a possible misbehaviour in dnsmasq/pihole-FTL at this point, but I'd consider that rather unlikely. If caching wouldn't work, we'd see far more reports, and I am also unable to reproduce your issue with my bare metal or Docker testing platforms.
If it is a bug, then the first hard part will be to reproduce it.
Unfortunately, I do not run LXC/proxmox myself. 
You could configure pihole-FTL to log some more details about how it's processing queries by adding the following flags to your /etc/pihole/pihole-FTL.conf:
DEBUG_QUERIES=true
DEBUG_EXTRA=true
DEBUG_EDNS0=true
Run pihole restartdns to apply those settings.