I think it will depend on what the shared understanding at the table is as to genre/tolerance, and also how much complication the GM is wanting to introduce into the situation prior to the players engaging it via their PCs.In this approach, still using jameson's example setup, would the guard let the PC in or escalate? Is making the real Diplomat get there before them part of the hijinks (as a complication to what the PCs were trying to do)? Or would that have been left out?
So the real diplomat could be introduced from the get go, to raise the stakes and ratchet up the tension, or it could be introduced as a response to a failed Bluff check ("Although you speak very sweetly and plausibly, it turns out that . . .").
But I think, on this approach, provided that it is genre credible that the PCs can get in, then the GM has to ultimately permit success.
When playing in this fashion, I find that one of the big challenges for a GM is deciding how much to introduce as complication/stakes from the get-go, and how much to leave as a response to players failing rolls. And there are at least two considerations in play here: if you introduce too many complications into the starting scenario, you run the risk of having no good material left when it comes to actually adjudicating the action resolution mechanics; and if your judgement of the complications that will work is out of synch with your players', then you can create anticlimaxes ("That was easier than we thought it would be!") or bogging down ("Can't we just get to the king already!") where you didn't intended to.
Personally, I would normally rather risk anticlimax than bogging down, and so would only introduce the diplomat as an initial complication if the players already knew that their was a diplomat in the offing, and were thinking about how to handle the issue, and hence already had some personal investment in a situation involving the diplomat.