If I recall correctly (its been forever), the glass face of the KP appears sort of like a super large watch movement; in that the contact mechanism appears to rotate as it "picks" and drops away. I've used magnetic stick versions of those (somewhere around the time Methuselah was born), They came in several versions. But I do not recall ever using a KP in a vital safety circuit, such as one of these "headblock" sticks.

On most of them, if not all, their front contacts could weld together, causing a less restrictive indication than intended. Thus they could not be considered fail safe. Again, if I recall correctly, their use was limited to non-safety related circuits, such as reception and transmission of the CTC code line from the office CTC machine, or to latch various non-vital controls and indications.

Big beefy mag sticks designed for safety related circuits were available, and more recently in a GRS B2 style plugin case. US&S also made similar relays. Since these were actuated by command from the office CTC machine (after safety checks), they would not usually be used at an intermediate signal, where no CTC controls or indications are available. Regular APBS sticks at the intermediates would be influenced though, by the headblock stick, via the HD line circuit or the equivalent indication from coded track circuits. They would then work the same magic at the intermediate as at the headblock.

Generally, directional sticks would determine which end transmitted the main code rate down the rails, and which end received it. As I mentioned in an earlier post. The receiving end also sends a code pulse, called the inverse code. But it is a very short pulse (less than 10% duty cycle), asserted on the rails only when the primary code pulses are off. The transmit end of each track circuit listens for these when it is not asserting the code rate pulse (50% of the time).

At each signal, headblock or intermediate, the main transmitted code rate is determined by the aspect displayed by the signal, and any other necessary considerations. It is then sent down the rails to the next signal, where it helps determine the aspect selected there. It can also be picked up by a train in the block and used for cab signalling and automatic train controls of any sort. The code rate selection at the next signal again follows the that signal's aspect. The process is then repeated again and again until the next siding is reached.

The inverse codes of necessity, follow the code rate of the main forward code. And It could therefore not show additional meaning by differing code rates. It could only be present or not present; the traffic section to the next headblock is clear, or it is not clear. This limitation is of course, the reason that the direction of the main code rate had to be switched back and forth, as trains went back and forth.

More recently, with the advent of hardwired audio frequency track circuits, It is now possible to send differing code rates from each end simultaneously and independently. As a matter of fact, the higher speed areas of the NEC, use multiple audio frequencies to send multiple code rates in the same direction, permitting a great many additional cab signal aspects, used for the higher speed authorizations presented to the Acela trains. Conventional trains still pick up only the standard set of codes on the first audio frequency.

When you used this bidirectional coding scheme, no line wires at all were needed for safety circuits, and with the advent ATCS and solar panels, no line wires at all! But one might still visualize the same "tumble down" and "tumble up actions here, from signal to signal to signal, so characteristic of APBS block signaling. This still happens with even the most modern of CTC systems. After all, nearly all CTC system designs were based upon APBS. To turn an existing APBS system into a CTC system, one could leave the track circuits and sticks mostly as they were, adding only interlocked OS tracks and powered switches at each siding end, and adding a CTC code line, now almost all via VHF data radio (ATCS)

As I recall, the WP signal training manual also showed a scheme where all this directional switching of the headblock sticks, was somehow automated, eliminating the need for the dispatcher to first set the direction, before he could clear a signal. But at the moment, it escapes me how that worked. Being retired now, it has indeed been forever!