Questions on the evolution of the GISS temperature product

Blink comparator of GISS USA temperature anomaly – h/t to Zapruder

The last time I checked, the earth does not retroactively change it’s near surface temperature.

True, all data sets go through some corrections, such as the recent change RSS made to improve the quality of the satellite record which consists of a number of satellite spliced together. However, in the case of the near surface temperature record, we have many long period stations than span the majority of the time period shown above, and they have already been adjusted for TOBS, SHAP, FILNET etc by NOAA prior to being distributed for use by organizations like GISS. These adjustments add mostly a positive bias.

In the recent data replication fiasco, GISS blames NOAA for providing flawed data rather than their failure to catch the repeated data from September to October. In that case they are correct that the issue arose with NOAA, but in business when you are the supplier of a product, most savvy businessmen take a “the buck stops here” approach when it comes to correcting a product flaw, rather than blaming the supplier. GISS provides a product for public consumption worldwide, so it seems to me that they should pony up to taking responsibility for errors that appear in their own product.

In the case above, what could be the explanation for the product changing?