(this should probably be a comment, but its a bit long for the wee box).
I don't know if it is possible. If it is possible today AND NOT DOCUMENTED then it might not be possible tomorrow. IMHO if this works, then it's potentially a defect.
When I previously had to solve the issue of provisioning with certbot across a cluster, I chose a different path.
Since you need to designate a specific node to run the initial provisioning and refreshes from (unless you happen to have some cluster cron in place) I chose to proxy that location on the designated machine on the other nodes.
An additional complication was that I was replicating the config from this master node. I was using nginx, but the approach can be applied with Apache httpd and probably other webservers.
I had 2 files in /etc/nginx/snippets describing how to handle /.well-known/. One was a proxy config, one was an origin config. These were replicated across all nodes. There was also a symbolic link at /etc/nginx/snippets/certbot.conf This link pointed to a file outside the replication tree. That file was also a symbolic link pointing back to one of the two initial files. On the master node, it pointed to the origin server, elsewhere to the proxy definition.
In my /etc/nginx/sites-enabled/*.conf files I simply included /etc/nginx/snippets/certbot.conf