GCP (grade crossing predictors) do work fairly well. But they are sometimes finicky, especially near block boundaries or in overlapping configurations over successive urban cross streets. And the underlying technology is not generally considered failsafe either. Therefore, they often are placed in redundant configurations, and have checking circuits to detect hazardous failure modes, any one of which which will summarily drop the gates if a fault is detected. In block signal or CTC territory they are also generally set up with a "Wrap around" circuit, using the entire block occupancy to activate the gates when it is occupied- instead of the failed internal electronics during a detected failure. This really extends the gate down time, annoying motorists a lot. But that only happens when there is a fault, and only when there is a train nearby.

When WP tried to put GCPs in next to the BART aerials, they had years of fuss and bother, because of stray BART DC propulsion currents saturating WP,s GCP coupling transformers. They really are finicky. Things eventually quieted down, but it never fully went away.

And yes, Sacramento RT does have automatic block signals in most off street areas. A friend of mine actually led the installation team. And they too had a number of "Teething problems". but I don't remember him ever mentioning any loss of shunt problems. Though they did have a lot of issues with excessive propulsion current saturating the track impedance bonds, causing false occupancies.

As to British shunting studies; one of the things they concluded, was that the use of cylindrical wheels (BART for example), intended to reduce truck hunting for a smoother ride, was very problematic. Turns out that truck hunting is essential to keeping the wheel and rail clean of rust, crud, and other dielectric materials.

Also noted were the dielectric properties of the wheel-axle set; some combinations of dis-similar metals in wheels and axles being unable to present a sufficiently conductive shunt (again BART for example) to reliably activate all but the very best of track circuit designs. Sacramento's, being two phase AC Vane Relay track circuits, were the best, for detection reliability at least. And BART's original track circuits being the worst (again - a perfect storm of poor shunting causes for BART).

Also of concern in British studies, and in a local study I was once involved with, is the effect on shunting reliability presented by none solid axles, sometimes found on heavy track maintenance machines, highrail equipment, geometry and rail detector cars, and of highest concern, low floor LRT cars where there is no press fit transverse axle at all. None of these do well at providing a high conductivity shunt, Hence lower quality track circuit designs, or those stretched to the limit in application, could not detect these axle sets very reliably. Reliable meaning 1/10^-9 or better failure to detect rate.


Seems nothing the human being does, will ever be perfect, even though it usually works fine when kept within its natural practical limits. I was therefore never one to push the envelope to the "cutting edge", also known as the "bleeding edge" in some circles. That is where people might bleed even when you don't want them to.