There's one part of this "the 'Close Door' button is disconnected" legend that really bothers me: the purported behavior is so trivially easy to test, but we keep falling back on "Otis Elevator engineers confirmed the fact", even though this is precisely the type of appeal to authority that we are all so quick to condemn when we observe it elsewhere. Several commenters have pointed out that they don't see this behavior in the elevators they encounter - so isn't it about time that we all did some rigorous scientific analysis?
Here, I'll start.
My own experience suggests that the close button often works, so that's the hypothesis I'm going to test. The elevator in my building is a Kone (unfortunately, I have no other information about it - no serial number was listed, and the State of Georgia doesn't seem to post elevator inspection details online).
After the doors first opened and I walked in, I observed a roughly 5 second delay on my stopwatch before the doors attempted to close. The same delay-before-closing was present when I called an elevator but did not step on, and when I took the elevator to a different floor, whether or not I stepped off. The delay before closing was reduced to three seconds for subsequent closing attempts if I interrupted the first closing of the doors with my arm. These measurements served as my baseline for subsequent testing.
I began a fresh test by stepping out of the elevator, letting the door close, and then calling it again. Upon entering the elevator, I immediately pressing the "Door Close" button without first selecting a floor, and observed the door closing immediately after. The same behavior occurred when I selected a floor before pressing "door close", but no change from the baseline was observed when selecting a floor without pressing "door close".
Conclusion: For this particular elevator model, the "door close" button does indeed cause the doors to close sooner.
I don't have a way of quickly determining whether there are, in fact, elevators out there that have intentionally disabled close buttons, but I've got a working theory about where this legend is coming from.
First, every time I've heard the claim that the "Door Close" button doesn't work, it has come from an Otis Elevator representative. It's quite possible that this is a claim that is only true for Otis elevators, but is only reported because there's very little news in putting on a ThyssenKrupp representative saying, "Our 'Door Close' buttons actually work!"
Second, I have been in elevators where selecting a floor would automatically trigger a door close event. It's plain to see that with this design, a door close button is redundant - but it's also easy to imagine a customer refusing to buy an elevator without a "Door Close" button. Adding a nonfunctional button allows the sales team to get that extra checkmark on the feature list, and also makes for a great story about "dumb management decisions" for the engineers to pass around.
I'd encourage you all to experiment with this on your own to see if this also applies for other manufactures. If you e-mail me your observations (firstname.lastname@example.org), I'll tabulate the data and provide it to anyone who is interested. I recommend the following test rubric:
1) How long does the door take to close when no buttons are pressed?
2) Does the time for the door to close decrease when a previous close attempt has been interrupted?
3) Does "door close" cause this time to decrease when no floor is selected?
4) Does selecting a floor cause this time to decrease?
5) Does pressing "door close" with a floor selected change anything?