Each is stored in a per-user file based on the job-specific dir name passed
to the condor.pm script located in the log directory named in
$GLOBUS_LOCATION/etc/globus-condor.conf file

Times to keep track of:
start_timestamp: earliest event we are willing to process
sweep_timestamp: latest event we are willing to process
flush_timestamp: recovery time which the container will use when it is restarted


Periodically:
    update sweep_timestamp to the current time

    if flush_timestamp has changed, 
        set remove_old_files flag to true

    For each file in dir:
        if (file is newer than start_timestamp) or (remove_old_files == TRUE)
            for each event in file:
                if remove_old_files is true
                    count submit and termination events

                if (older than start_timestamp) or (newer than sweep_timestamp)
                    ignore event
                else
                    queue event

        if (file is older than flush_timestamp) and (#submits = #terminates)
            remove log file

    sort events from all files based on time
    send SEG events to container

    update start_timestamp to be sweep_timestamp