Hot standby failing with page # of relation # is uninitialized

Debian Stretch AMD-64, Postgresql 9.6.3

Streaming data from live server and we see these messages on the hot
standby server.

2017-09-06 14:42:09 BST    11548 WARNING:  page 30558 of relation
base/2497261/2504373 is uninitialized
2017-09-06 14:42:09 BST    11548 CONTEXT:  xlog redo at 44/5E0B7CA8 for
Heap2/VISIBLE: cutoff xid 29182314 flags 1
2017-09-06 14:42:09 BST    11548 PANIC:  WAL contains references to
invalid pages
2017-09-06 14:42:09 BST    11548 CONTEXT:  xlog redo at 44/5E0B7CA8 for
Heap2/VISIBLE: cutoff xid 29182314 flags 1
2017-09-06 14:42:09 BST    11545 LOG:  startup process (PID 11548) was
terminated by signal 6: Aborted
2017-09-06 14:42:09 BST    11545 LOG:  terminating any other active
server processes

Have to resync the files on the hot standby, restart the standby and
streaming continues but then stops again a few days later.

No obvious data corruptions as pg_dump generates backups on the effected
databases and users are accessing the live server without issue.
I think it is a timing issue in the generation of the WAL's as if we run
a vacuum/analyse/pg_dump on the live server against the effected tables
no errors/issues are reported in the log.

Any help gratefully received.

Adrian Heath
*TrackTrans Limited*