As with many things in chess, the answer is: It depends on the position. You touched on this by discussing open vs closed positions, but it can go much deeper than that. In high level play, the middle game may be closed, but the bishops may be preferred as part of an anticipated endgame strategy.
Knights are superior to bishops in more aspects than just closed vs. open positions. A knight is generally the better option when it comes to blocking an opponent's passed pawn, for instance.
The biggest differences come in endgame strategy, especially when there is only 1 bishop or knight. There may be two knights vs. two bishops at some point in the middle game, but that can easily change as part of a forced exchange, with the resulting position being that the remaining bishop is poorly suited for the emerging endgame position.
In the case of two knights vs a rook -- it depends on the position. Two well placed knights, or knights that are participating in a plan, are going to be better than an inactive, semi-active, poorly placed rook -- open position or not.
The knights also do not need to be linked in order to be effective. There are certainly situations where it is useful for them to be able to cover each other, but there are disadvantages to this as well -- namely that they will limit each others mobility.
With all other things being equal, the bishop pair is generally superior to a knight pair. However, all things are very rarely equal, and taking maximum advantage of the positional imbalances is a cornerstone of any good chess strategy.
IM Jeremy Silman has some excellent books on learning how to do this.
Here are some articles you might find useful:
Bishop vs. Knight in endgames