That's not exactly hard to do, depending on features. There's a definite trade-off between features and the type of regex engine that can be implemented.
It seems to me that one should be able to detect which features are actually used in a regular expression and choose one of multiple different underlying implementations based on that.