I am looking for examples of scenarios where the believed-best moves among experts were proven wrong by a computer.
This might not be quite what you're after, but computer-generated endgame tablebases have revealed new knowledge that includes overturning previously held beliefs about the evaluation of certain endgames. This means that the experts were wrong about what the best moves were at plenty of points in such endgames. As an example, it's noted in the endgame study section of Wikipedia's tablebase article,
For instance, the endgame with a queen and bishop versus two rooks was thought to be a draw, but tablebases proved it to be a win for the queen and bishop, so almost all studies based on this endgame are unsound.
That article cites John Nunn's Secrets of Pawnless Endings for that fact, and Nunn's whole Secrets of ----- Endings series would probably be a good source for more examples of this sort of evaluation turnaround, as those books are the end result of Nunn's extended exploration of tablebases.