Streaming replication fails after some time with 'incorrect resourcemanager data checksum'

I already posted in the slack chat #help channel but got no answer :-(

We have a read only / hot standby system and are facing the same problem as described in https://stackoverflow.com/questions/35752389/incorrect-resource-manager-data-checksum-in-record-at-2-xyz-terminating-walrec (the post is already 3 years old).

That means after some time (sometimes two days, sometimes half a day), postgres starts logging 'incorrect resource manager data checksum in record at xyz' and shuts down wal receiver (and stops streaming replication).

Master and slave are running on Ubuntu 18.04, Postgres 12.1, ext4 file system (no zfs or btrfs, just lvm on the master); we only use ecc memory (192 gb on the master and 256 gb on the slave) and nvme ssds on both servers using a soft raid 1.

When the error occurs, a restart of postgres on the slave "fixes" the problem.

Any ideas what we can do to prevent/investigate the problem?

Kind regards

Julian