The appearance of a well-known character - particularly a villain - provokes a strong response from the audience and an anticipation for what their appearance means for the story. Used carefully, this can be a good shortcut to investment in the proceedings but it also runs the risk of over time turning established characters into caricature of themselves. The simplest illustration of this is with characters from sketch shows such as the Fast Show. They appear on screen and the audience knows that before long they are going to spout their catchphrase at which point the viewer will laugh. It is an almost Pavlovian response - a release of the anticipation regardless of how funny the preceding sketch may have been and it erodes the value of the character by robbing them of any depth.
In a story the effect is less obvious, but potentially more of a problem. Part of a dramatic scene is the tension of not knowing how events are going to unfold and a known character can very quickly bleed any uncertainty away. The characters may be in danger, but of course The Hero is going to survive. The scene is no longer about whether the characters will survive; it is about how they will survive. If the character has a history of surviving by being cocky and lucky then that question is answered too and the audience is left waiting for the action to go through the usual motions.
That isn't to say this is a bad thing. The A Team, for example, survived through five seasons by rarely deviating from an identical formula (get job, meet badguys, be captured, build implausible device, win). In Castle you can usually identify the murderer in any given episode by noting the character introduced in a particular (small) time window. The episodes of shows like these have a familiarity which makes them very easy to watch however they buy that familiarity by discarding any genuine tension and have to work hard in other areas to ensure they continue to draw in the audience despite being increasingly predictable.
A roleplaying game needs to earn the players' attention by giving them the opportunity to succeed at something and overcome an obstacle. The simplest way to generate tension is by putting the characters in some kind of danger which they can fight to overcome. While there is usually the unspoken agreement that the gamesmaster is not actively trying to kill the players (it is hardly a challenge to kill a PC when you control every last variable in a game world) it is important for everyone to believe a total party kill is a possible outcome to an encounter, however unlikely (and undesirable) else the challenge is robbed of its teeth. Equally, the players need to be able to win. While the definition of "win" is very variable depending on the game and the story, to have value it needs to be earned through the players' decisions and actions rather than some Deus Ex Machina.
The exception to this is the occasional dramatic encounter, when the odds can be completely out of balance and the world is rigged to ensure a particular outcome. This device needs to be used sparingly as it is non-interactive which doesn't fit well with the most basic premise of a roleplaying game as the players' story and it needs to be pretty obvious what is going on so it doesn't look like the GM has intervened to fix a botched normal encounter.
This brings us back to the characters - and particularly the villains. In a recent Star Wars game I had a Dark Jedi (Inquisitor Ceres) turn up with all the usual Imperial pomp and ceremony. She scared the living daylights out of the players because I managed to balance the known (she was described in a very Vader-esque way and happily Force Choked an important bureaucrat, demonstrating her own power on a number of levels) with the unknown (she wasn't actually Vader).
For this mission I had considered using Vader himself. It would have been an easy switch but I think it would have detracted from the situation as the players (all very capable and experienced gamers) would have inevitably calculated his involvement in the forthcoming scenes. They were low level so they knew Vader was well beyond anything they could sensibly face in a real encounter therefore either he wasn't going to confront them directly (and his appearance on-world was a cheap scare tactic), or he would show up in a dramatic encounter where they weren't in any real risk of being killed by him. By using an unknown Dark Jedi I removed this safety net and generated fear because they didn't how powerful she was and therefore couldn't guess her exact role in the story. Basically (in the words of the Jedi player) they might have to face her.
The mission ended with the evacuation of a Rebel base as the players held back a horde of stormtroopers. Despite the stormtroopers being about as effective as they were in the films (not my finest moment) the players still chose to withdraw and flee for space when Ceres turned up which made for a far cooler ending than if I had forced them to leave by putting an insurmountable obstacle in their way (if Vader had occupied the same role).
I wish I could say I'd thought all this through in advance, but I was lucky. The effect was pronounced though - the players spent the mission operating under a sense of urgency because the Empire was coming and when it arrived it would be under the command of a terrifying Force-wielding maniac who would be actively gunning for them. The character list is an important part of the gamesmaster's toolset but learning how exactly those tools can be used for different effects is fascinating.