Thank you both. I'm working on this. Since this will generate lots of data, I'm changing my test case to stop right after the overlay mount to avoid a bunch of "node configuration" steps. This should get us a cleaner test case without filling the logs and dumps. I'll update the config file WRT ACL. I'll get back to you soon.
+1. yes. tcpdump of both nfs and gluster traffic (taken on the node where nfs-ganesha server is running) may help . Also can you please check for any errors in gluster brick logs and ganesha-gfapi.log as well.
One difference between nfs-ganesha and Gluster-NFS is that gluster-NFS does all the operations using root user, where as nfs-ganesha/fsal-gluster switches to nfs client user creds before sending requests to Gluster.
Also I noticed in the logs that there are some calls to fetch ACLs even though you seem to have disabled ACLs using global option. If there is no need of ACLs, could you disable it in EXPORT block {} instead and give a try.