Some programs (NFS4 and SLURM) see BTRFS filesystem empty

[merilius]

I have a fairly large (49TB) BTRFS file system which I was successfully exporting via NFS to another machine.
After the last system update and reboot an NFS4 export of a BTRFS filesystem results in empty directory. There was no configuration change.
I.e. mount succeeds, but an empty directory is a result. Moreover listing a this directory on the NFS server from within SLURM (queue manager) reports an empty directory as well. On the other hand no problems listing it from Bash, Perl, Python, R or Midnight Commander.

I FSCK'ed this filesystem and there are no errors.
Exporting an ext4 filesystem works without issues.

Debian Stretch, all up to date.
btrfs-tools 4.13.3.-1
Kernel: 4.14.0-2-amd64 #1 SMP Debian 4.14.7-1 (2017-12-22) x86_64 GNU/Linux

# btrfs check /dev/mapper/fc_trunk-part3
Checking filesystem on /dev/mapper/fc_trunk-part3
UUID: 40a2e65b-f34a-4d33-946d-055d93fe7ffa
checking extents
checking free space cache
checking fs roots
checking csums
checking root refs
found 17984968683520 bytes used, no error found
total csum bytes: 17528852232
total tree bytes: 31246974976
total fs tree bytes: 5960253440
total extent tree bytes: 6075351040
btree space waste bytes: 1894287388
file data blocks allocated: 72243434635264
referenced 24376557264896

[merilius]

I think what might have happened:
In Systemd mounting of a filesystem is executed early, but the system continues to boot.
Mounting this filesystem takes a good 20 seconds, so SLURM and NFS kernel server might have been executed during the mounting, and both processes accessed the BTRFS partition immediately resulting in a filehandle to an empty directory (mount point).
While restarting SLURM is not a big issue, NFS runs in kernel, so systemctl restart nfs-server is not enough.