"nfs server not responding, timed out" with Linux client and 13.0-RELEASE server

From: Alan Somers <asomers_at_freebsd.org>
Date: Tue, 01 Jun 2021 21:20:05 UTC
I have several Linux NFS clients (Debian, with various 4.9 kernels)
mounting NFS 4.0 from both Solaris and FreeBSD (mostly 12.2-RELEASE)
servers.  I frequently see kernel messages in the clients about "nfs:
server <HOSTNAME> not responding, timed out" for both the FreeBSD and
Solaris servers.  It usually recovers after a few seconds.  However, on the
sole 13.0-RELEASE server, the hangs last for minutes to hours.

tcpdump/wireshark shows the NFS client sending a "V4 Call SEQUENCE" packet
every 5-10 seconds, all with the same sessionid.  The server responds with
a TCP ACK, but sends no NFS-level response.  I've seen this with both NFS
4.0 and 4.2.

NFS delegations are disabled, and nfsdumpstate shows that the client has
callbacks enabled, with 2-4 opens, openowners, locks, and lockowners.

Can anybody suggest how to fix or further debug this problem?