Models seem to go through periods where from run to run, there are significant changes in medium- and even short-range NWP model forecasts that have profound impacts on the end-users' decision-making with respect to the future weather in their WFO area of responsibility. Often, the reason given for the forecast change is that new data got into the data assimiliation system that resulted in the forecast going in a different direction from before. This is described conceptually in the COMET data assimilation system training module (click on DA Process from the menu).
However, the reason likely has much more to do with the inherent predictability of the atmosphere from a given initial condition than the NWP model getting "hammered" with new, radically different data. After all, data assimilation systems generally start from the premise that the model's short range forecast, used as the first guess by data assimilation systems, is of good quality, thus having a strong influence on the new analysis.
The first example following this introduction illustrates this point.