This is the old creationist argument "scientists use layers to date fossils, and fossils to date layers. It's circular!!!!" argument all over again. Someone has partially grasped what is actually a very common, well-supported and well-recognised approach in the scientific community, and excitedly rushed off to trumpet to the world that she's exposed the "fatal flaw" in those frauds' work.
What she's getting all in a lather over is, in effect, the distinction between *precision* and *accuracy*. Precision is a measure of how closely an instrument repeats the same measurement under the same conditions; accuracy reflects how close that measurement is to the *true* value.
Now, it's been possible to make thermometers that are very, very *precise* for a long, long time - glass is very rigid and doesn't change shape, and the thermal expansion properties of mercury don't change with time. But *accuracy* is another question. Old thermometers were, of course, hand-made, which included the etching of the temperature marks. Even a sub-millimetre error in alignment will lead to a systematic error on the order of a degree.
So what happened at Amberley in 1980? Well, who knows. The timing seems right for an upgrade from a mercury thermometer to an electronic thermocouple - but I'm just guessing. The step change there *could* be a real meteorological phenomenon (a very strange one indeed). So how do you check? Well, you do what the BOM did: you look to see if a similar sudden change occurs in the nearest surrounding stations (it doesn't). That points to a systematic error that needs to be fixed; since the culprit is almost certainly a change to a measuring device with a different systematic error (note: *all* measuring devices have a systematic error), the obvious solution is to apply a constant offset to the older data to bring both instruments onto the same footing.
Marohasy seems to call this fraud. Those less prone to flights of fancy would call it "calibration".
