There is a static model, and one you exchange to during the animation sequence? That one would be quite high poly?
The polygon count is the same but, the static version, being static, is very carefully optimized with multiple LOD levels, and it has no bones. This is what makes it easier on fps than the animated version, more than the bare polygonal count.
But if we tried to use the same model as animated, we couldn't optimized it so much, because each LOD level would have to have an active animated skin, otherwise jetways in the distance would snap out of the AI door, basically looking as they are now.
How about only exchanging the one for the activated gate (my own plane) to the high poly model during the animation?
Not possible, because when the jetway is docked to an AI plane, is STILL "animated", even if it doesn't move, because its final position depends on the location of the AI doors, so it's different for every parking, meaning we cannot simply revert to a static model when the animation is done but instead, it would require keeping the animated (slower, even when it doesn't move), we would have to keep the skinned+animated version for all the time the jetway needs to be in a docked state.