How would the data produced by the Goulburn AWOS be "found" to be erroneous or anomalous by "professional users"?
A "professional user" (your quotes) may be a researcher, an air conditioning designer, a farmer or one of the myriad data users. It may be one of the private weather agencies. Their suspicions may be aroused enough by an extremely low temperature for them to query it. Or they may have significantly different nearby data from a private station (lots of them around). In any case, a query raised by the user would be checked.
Who, precisely, using what measuring equipment, precisely, made what measurements to make the finding, when precisely?
Ask the bureau, I'm sure they'll answer your question.
I'd make a wild guess that no one in BOM decides.
And like most wild guesses, it would be wrong.
BOM's purchased some magic bean software that was promised to do the BOM's work for it. And that software has been programmed to pretend to do what used to be done through tedious manual checking by human professionals.
If you expect the data checking to be done manually you're dreaming. There just aren't enough people who can do the checking properly, and even if there were, who would pay them? I certainly wouldn't because it would be a completely inefficient way of checking. As one who did this sort of checking manually back in the day, I can assure you that computer checking is far more effective. The "magic bean" software is Australian written, and is arguably the best in the world. I say that as one who has used it since its inception, maybe 30 years ago until my final retirement a few years ago. Its design is guided very much by the users - the BoM plus virtually all Australian water authorities, some mining companies and a few others - who test the daylights out of it and also specify particular things they want it to do.
It does have range checking, and sure, someone would have set the limits. The limits would be different for each station, or stations in a region, and set by the person/people responsible for that station. The person who wrote the program wouldn't know or care what a reasonable value was, because it's not hard coded. I don't know about the BoM, but normally anything outside the expected range would raise a flag for the data to be checked. I've told you in my previous post what that could result in.
I don't carry any particular candle for the BoM, and have never worked for them, but as I've said, I've had a bit to do with them as both a professional user and data supplier. Sure, occasionally errors will get through. You may never have made a mistake, but most of us aren't that good. To say that because one piece of data is wrong, all data is suspect, is a load of rubbish.