I agree Wouter. DCC only operation should be selected in CV29 such that absence of DCC commands being received (ie - DC rather than DCC on the track) causes the decoder to decel to zero speed. the decel rate (momentum) and therefore the stopping distance is programmable. So is the delay on recognizing there are no DCC commands being received. Every decoder brand may be different. Some work better than others.
There are always some conditions, limitations and exceptions that come into play when thinking about unsupervised automation. Ask the guys using a kill switch to protect an open bridge about using Keep-Alive :)
If all else fails use an operator/engineer per train that is always paying attention.
---In WiringForDCC@..., <vandoornw@...> wrote :
Killing power does not work as instantaneously as you might otherwise expect in case of keep-alive batteries! If there is, there's also momentum to consider.
What DCC lacks here, is the option for a computer to steal control from a user instantly and unconditionally. But that would open up a whole new can of worms...