Some general information:
EB platform
EH900 would appear to work as follows (I may be wrong!):
Normal signal indications:
EH9 red, EH900 red
Approach-controlled 1:
EH9 red. EH900 red
EH900 goes to yellow as the train passes the intermediate home. EH9 remains at red
Speed controlled:
EH9 red. EH900 red
As the train passes the inner home, it starts a timer. EH900 goes to yellow when the time ends and EH9 remains at red.
(A standard timer may be 4.5 seconds. If a train reaches EH900 before then, the signal will still be red and the train will get tripped)
Approach-controlled 2
EH9 red. EH900 red
EH900 goes to green as the train passes the intermediate home. EH9 goes to green
Manually controlled:
EH9 (starter) green, (and so) EH900 green
Providing that EH9 is not being held at danger (regulated), it is assumed that the usual operation would be:
With no train in the overlap of starter EH9, EH900 will operate as per approach-controlled 2.
If there is a train in the overlap of EH9, depending on the position of that train, Speed controlled will be used
If EH9 is being regulated and there is no train in the overlap of EH9, Approach-controlled 1 will be used.
WB platform
It would appear that the trainstop is just the standard speed-controlled ‘policeman’. Based on the usual timer, the trainstop will lower after a preset time. If the train is going too fast, it will get tripped at the trainstop.
Common uses of policemen usually for enforcing a speed – such as running into a terminus platform / siding, or as the equivalent of the last inner home signal, but without the aspect. One was used in Highgate NB platform and also Waterloo and Oval NB platforms as part of the speed signalling to allow trains to get closer together by cutting down the necessary overlap.
I’m not sure what the specific reason is for the one at Embankment. However, the reason for the difference between EB and WB arrangements is probably due to there being a crossover east of the platform and, I assume, the EB starter is also used for regulating.
I'm sure that the single trainstop is the preferred method where possible, because of it's cheapness
From a driver's point of view, getting tripped at it shouldn't count it as a SPAD (despite some managers in the past thinking that it was) because drivers aren't supposed to be keeping their eyes on the track looking for trainstops!
In the end, I suppose that it's horses for courses. Using a draw-up signal gives more flexibility than a policeman as can be seen from the above information.