pg_replication_lag_seconds seems to "freeze" in 0.13.2

[megamail2000]

The replication metrics request (taken from pg_replication.go) returns an actual replication state:

postgres=# SELECT
CASE
WHEN NOT pg_is_in_recovery() THEN 0
                WHEN pg_last_wal_receive_lsn () = pg_last_wal_replay_lsn () THEN 0
ELSE GREATEST (0, EXTRACT(EPOCH FROM (now() - pg_last_xact_replay_timestamp())))
END AS lag,
CASE
WHEN pg_is_in_recovery() THEN 1
ELSE 0
END as is_replica
;
 lag | is_replica
-----+------------
   0 |          1

While the exporter still outputs the maximum value from the past:

# curl -s http://10.6.251.206:9187/metrics | grep pg_replication_lag
# HELP pg_replication_lag_seconds Replication lag behind master in seconds
# TYPE pg_replication_lag_seconds gauge
pg_replication_lag_seconds 1.138475469918e+06

In this case the other metrics are updated properly.
How can I reset the output with an actual zero lag value?

Environment

• System information:

  Linux 5.14.21-at154.2.2-default x86_64

• postgres_exporter version:

  0.13.2

• postgres_exporter flags:

DATA_SOURCE_NAME="user=postgres host=/tmp/ sslmode=disable"

• PostgreSQL version:

  psql (PostgreSQL) 15.3

[sysadmind]

I think #895 should fix this. It will be in the next release. If this does not resolve the issue, feel free to re-open.