archive_lag_target is meant for forcing a log switch after x amount of time. A value of archive_lag_target=0 means you've disabled time based forced log switch.
In your case every checpoint fired at 3 secs is facing this problem. This could indicate you've slower IO subsystem.
Meanwhile what filesystemio_options you are using?
Initially I read logfile size as 500 m instead 50m, I agree with John on increasing logfile size as well to avoid agressive log switches.