“That tiny difference was well within uncertainties. There is no statistical difference between adjusting the temperatures recorded by those poorly-sited stations versus discarding all data from those stations.”

This goes to show a key fault of denialistic thinking: favoring the fact of existence of a confounding variable (here, the thermometer position) over quantification of that confounding variable’s actual effect on the conclusion, esp. when such quantification can be and actually has been done.