The theory that different areas of the tongue tasted different things - the Tongue Map - has been pretty thoroughly debunked but lived for aboutt a century. This seems like something fairly easily testable.

Bullet lead analysis gained scientific acceptance and stuck around for forty years; it still is viewed as good science by many, although its probative value may have been overstated.

Bruise aging was accepted for a shorter period of time, but appears almost worthless. This was another testable hypothesis that lasted longer than it should have.

I don't have particularly smart things to say about why these errors lasted while others were destroyed by truth. Perhaps someone else does.