There is a flight shortly flying YMML to 7752S16707E and then back to YMML.
Neat that one of the waypoints is named BYRRD.
Tracked for about 75 minutes after takeoff and now sat on an “estimated” position that is slowly drifting back to the airport.
In reality, it is still heading towards Antarctica and is not expected back in tracking range for ten or more hours.
Three and half hours after takeoff onwards, aircraft tracking shows it to be hovering over the originating airport and flagged as an “estimate” when in reality it is more than half way to Antarctica.
Flight has now reappeared on the tracking and is on the final leg of the journey back to YMML.
And landed at YMML about half an hour early.
Last time around this was because of bad data from Airservices Australia (a bad ETA); I expect it’s the same here.
No, this time the landing data was fine and agrees with the land-based ADS-B tracking data found in the public track log.
The only issue was the estimated position hovering over the airport for at least 8 hours when it was, in reality, several thousand miles away.
The estimated position will be what is derived from the ETA (Estimated Time of Arrival), which FA gets from Airservices Australia.
Maybe the time wasn’t UTC 
As soon as FA has ADS-B coverage, data from Airservices Australia should be ignored.
In this case, the departure and destination airport are the same and the track out to Antarctica and back is a single line, not a loop. This causes the estimated position software to fall over as it does not traverse the flight plan positions.
Do they descend for the Antarctic coastline?
This was a sightseeing flight i presume?
The actual landing time is not the same thing as the ETA.
This is consistent with having a bad ETA that claims that the flight should have already arrived at the destination. Given the data available to the public version of the flight while it was being estimated mid-flight (the last terrestrial position, the expected destination, and an ETA in the past), there’s not really anything more reasonable to do here. The change since the last flight was that we no longer synthetically arrive the public version when we have other versions of the flight that are definitely in the air, so you don’t see a premature arrival before it flies back into terrestrial coverage.
I assumed it had something to do with the fact that the journey started and ended at the same airport, so an average position was still at the airport.
I guess these silly anomalies will vanish once there is more of the space-based ADS-B data being used.