TalkChess.com

[Evert Glebbeek]

Seems to me that there is a bit of confusion caused by two different ways to do the root search: either you call your normal search function, which returns both a score and a best move, or you play each move in the root function and then call your normal recursive search function.
In the first case, obviously you research all moves after a fail high/low, because you call the search function again with a modified alpha/beta window. In the second case you could indeed only research a move that failed high as long as you already have an exact score from an earlier move.

I've always done the latter myself.