One of my favorite examples is bite mark analysis. If you follow the references in studies, they all say that they point to experimental evidence, but the bottom just falls out and there's nothing but some claims that some guy thought it's possible.

The Obama administration actually started a National Commission on Forensic Science that made lots of great recommendations for how to fix most of the problems with forensic science. Of course, the courts and the cops didn't like that they had to throw away precedent for tools they could use to convict people, and the Trump administration dissolved the commission, so it got put on the back burner. Hopefully it will be taken up again soon.