It's such an old story that you'd have thought there would be an explanation online by now of exactly which telephone exchanges had this problem and when those telephone exchanges were in use.
For what it's worth, it didn't work when I tried it, probably in the 1980s. Perhaps it worked in the 1970s in some places?
I can't test as I've not had a voice capable land-line for some time. It may not work on newer exchange equipment. It won't work if you have a service whereby calls are directed over a digital connection. It has never worked for mobile phone services. It doesn't work on some (most? all?) office PBX arrangements, either.
As well as allowing this sort of scam to operate, the "feature" can also be used as a DoS attack, blocking calls to and from a line for a time.
1. Users do not expect the call to drop if for some reason the connection to the recipient momentarily is interrupted - there doesn't seem to be any obvious reason it should be right? Remember users don't understand how any of this actually works.
2. But, when they hang up, which to you as the exchange looks essentially the same as the connection being interrupted, they expected the call to end.
As an extreme example of (1) suppose you have a rotary telephone and idly while talking on a call you put a finger in the dial and dial a '4'. What do you expect to happen? Nothing right? Maybe it makes some click noises, and you apologise to the person at the other end. But those click noises are the same - for a fraction of a second - as hanging up. So if we just naively code the system to drop calls whenever it thinks anybody has hung up, the call drops. You can bet customers will not be happy.
So the providers would pick a plausible delay. OK, let's say after two minutes with the call recipient showing as closed we'll give up and end the call, the caller can always just call back if that's a mistake anyway.
Well this fraud comes into the picture, adding a third consideration to the balancing act. Most UK providers would go in and choose a new smaller timeout. One second seems to cause false positives. How about five seconds?
So there's going to have been a window, and it will vary depending on where you live. System X was definitely not everywhere in the 1980s. It will also depend how you tested. If you tried waiting 5 minutes and the timer was set for 30 seconds then your test seems fine, but a victim who hangs up, counts to five and then tries their bank will get stung.