It’s my understanding that observations were taken at a few different locations, mostly in or near the city (the airport is nearly 20 miles west). Personally, I only compare apples to apples when looking at temperature data (1948-present), because there are all sorts of variables that could influence whether or not data from the 1880s at a different location could be compared to data from the 2020s at the current location.
As for Harrisburg, it seems to me that if a 30 year weighted average went up by 2 degrees by taking and replacing 1/3 of the data, it would stand to reason that the replaced data was 6 degrees lower than the new data. For example: suppose the averages for the 1980s, 1990s, and 2000s were all 70 degrees. 70+70+70 = 210 / 3 = 70. To get that average to 72, we would need the 210 to become 216, so 70+70+76 = 216 / 3 = 72. Am I missing something?
Regardless, even if the temperatures for a particular month somewhere were 3.5 degrees warmer (and not 6) than they were just 30 degrees earlier, it would and should set off major alarm bells.