The fail bit is harder indeed, especially for larger regexps, but not totally impossible. The easiest way towards that goal seems to be constructing the DFA fir

The fail bit is harder indeed, especially for larger regexps, but not totally impossible. The easiest way towards that goal seems to be constructing the DFA first and then generating illegal single edits (insertion, deletion, substitution). Generating a positive example is possible without it.