I wrote:
> I think that walsenders ought to report their current replication command
> as though it were a SQL command. They oughta set debug_query_string, too.
I also notice that a walsender does not maintain its ps status:
postgres 921109 100 0.0 32568 11880 ? Rs 18:57 0:51 postgres: publisher: walsender postgres [local] idle
I don't mind if we decide we don't need to reflect the walsender's
true activity, but let's not have it showing an outright lie.
regards, tom lane