[D] When to use SMOTE when dealing with rare events classification?

I'm reading that SMOTE is a common technique for the classification of imbalanced data. What could be the downsides of SMOTE and when is it useful?

classification events machinelearning smote