It warrants professionally-cautious skepticism because the synthetic data generation process often involves assumptions, approximations, and spurious or omitted dynamics. These limitations can then impact the model, as it is learning on dynamics present in the synthetic data but not real life.

Not saying there is no place for synthetic data. Just needs to be in situations where the salient dynamics are well understood such that they are realistically reproduced in the generated data.

The synthetic data is generated by a model, but if you need synthetic data to supplement empirical data for model estimation, isn't that because you don't have enough data to confidently say what the model for the data is?