We had an interesting occurrence in a playtest of my civ-building game the other night. The game lasts between 36 and 42 rounds, with a die roll at the end of each round determining whether the game ends or continues for (at least) one more round.
In our game, in the 36th round, a battle weakened my hold on one of my key territories. Had the game lasted another round, one player would have swooped in, taken control of the territory, and prevented me from scoring the points from the territory, and giving him the game. But, the die roll at the end of the 36th round ended the game, and I ended up winning.
I have been thinking about this from the perspective of what players might say about the game's random ending condition. I think there are two different schools of thought. On the one hand, someone could make the case that the random die roll decided the game; if it had been X, the other player would have won, but since it was Y, I ended up winning. Such a view might perceive this aspect of the game to be a flaw, because it lets the outcome of the game hang on something that is out of the players' control and is not known at the start of the game. On the other hand, it could be claimed that players do know for certain that they will get 36 actions, and can't and shouldn't count on more actions than that. Thus, leaving anything undone after the 36th round incurs risk. This view sees the emphasis placed on contigency planning, and requires players to cover their bases but have a plan for how to make extra points should extra actions be available, or perhaps even bank on the extra actions but accept the risk that this brings.
Personally, I find that the uncertain game ending is a big part of the enjoyment of the game for me and so I tend to view the game from the latter perspective. But I don't think the former perspective is entirely wrong, either -- it really comes down to a matter of taste. I also suspect that one's preferences may be influenced by the scope of the game that one is playing. I don't think anyone minds a healthy dose of randomness in a 20 minute game, but in a 3 hour game like this, I expect there will be some that will be very vocal in their disapproval of it. What I am uncertain of is whether there will be a significant quantity of players who actually prefer this feature, or whether my own preferences come through too strongly.
(I suppose that for players who are really unhappy with the variable ending, I could just add in a variant that brings the game to a hard ending at 36 actions. This certainly wouldn't break the game).
I just found it interesting as a case where there isn't a "right" or "wrong" answer from a design standpoint -- it's highly subjective. I welcome thoughts from others on experiences they've had with variable game lengths or game design issues where a mechanic was (or could potentially be) liked by some and loathed by others.