Setup: NFS server with either Leap 15.2 or Tumbleweed and a Linux client
mounting a directory from the server. There's no difference if NFSv3 or NFSv4
is used.

1. With OpenSUSE Leap 15.2 on the nfs server machine
Server:
  # uname -a
  Linux orion 5.3.18-lp152.78-default #1 SMP Tue Jun 1 14:53:21 UTC 2021
(556d823) x86_64 x86_64 x86_64 GNU/Linux

Client:
  # mount
  orion:/video on /var/spool/video.orion type nfs4
(rw,relatime,vers=4.0,rsize=1048576,wsize=1048576,namlen=255,soft,proto=tcp,port=0,timeo=100,retrans=2,sec=sys,clientaddr=192.168.1.21,local_lock=none,addr=192.168.1.17)
  # cd /var/spool/video.orion/dir
  # dd if=00001.ts of=/dev/null bs=1M
  2000+1 records in
  2000+1 records out
  2097399440 bytes (2.1 GB, 2.0 GiB) copied, 17.9036 s, 117 MB/s

No issue, file transfer succeeds


2. With OpenSUSE Tumbleweed on the nfs server machine
Server:
  # uname -a
  Linux orion 5.13.4-1-default #1 SMP Thu Jul 22 15:55:06 UTC 2021 (91a0cca)
x86_64 x86_64 x86_64 GNU/Linux

  Syslog:
  Aug 12 23:46:52 orion kernel: [52908.001097] rpc-srv/tcp: nfsd: sent 102156
when sending 581760 bytes - shutting down socket

Client:
  # mount
  orion:/video on /var/spool/video.orion type nfs4
(rw,relatime,vers=4.0,rsize=1048576,wsize=1048576,namlen=255,soft,proto=tcp,port=0,timeo=100,retrans=2,sec=sys,clientaddr=192.168.1.21,local_lock=none,addr=192.168.1.17)
  # cd /var/spool/video.orion/dir
  # dd if=00001.ts of=/dev/null bs=1M
  dd: error reading '00001.ts': Input/output error
  98+0 records in
  98+0 records out
  102760448 bytes (103 MB, 98 MiB) copied, 123.445 s, 832 kB/s

  Syslog:
  Aug 10 22:01:33 vdr kernel: [17893.504201] nfs: server orion not responding,
timed out
  Aug 10 22:01:36 vdr vdr: [1348] ERROR (tools.c,423): /var/spool/video:
Eingabe-/Ausgabefehler
  Aug 10 22:01:36 vdr kernel: [17896.128046] nfs: server orion not responding,
timed out
  Aug 10 22:02:02 vdr kernel: [17922.980069] nfs: server orion not responding,
timed out

The errors don't occur immediately, it works for some time before the nfs
timeouts are starting.
I've seen the rpc-srv/tcp error mesage in the syslog of the nfs server only
once, normally there are only the nfs timeouts on the client and no error
messages on the server.

The issue looks very similar to
https://bugzilla.kernel.org/show_bug.cgi?id=213887 .