Version 9.2 is very old and has a lot of issues related to streaming replication. Additionally, it is already EOL.
Agreed.
Furthermore, max_standby_streaming_delay has no relation with the streaming lag. It's associated with queries conflicting with data. This parameter will not speed up the WAL apply operation.
Yes, I know. But this parameter determines how long the standby server should wait before canceling standby queries that conflict with about-to-be-applied WAL entries. So, there could be lots of queries conflicting with about-to-be-applied WAL entries.
Again, I was not sure about what is causing the delay so I needed to try any possibility.
Kindly consider upgrading to a supported version.
Yes. We have a project in progress to upgrade.
On top of that, I can say that I have restarted the slave and the replication is now normal.
You can see above that the replication delay is not passing 3 minutes, which is the max_standby_streaming_delay, but the actual average is 5 seconds. So, looks like rebooting the EC2 instance resolved the issue, but I do not know why.