The diodes will set you back 2-cents a piece so why not just put them in. 
I agree with GRJ that the effect will be minimal and in all likelihood unnoticeable. Note that the coil is driven by a half-wave signal. So even when the coil is energized, at least half of the time there is no energy in the coil and hence no energy for the proposed new diode to re-circulate!
The capacitor could indeed add more delay, but as George points out you can hear the AC half-wave buzzing if the switch is jammed in-between - what amounts to a free audible warning/alarm. A large capacitor would smooth the AC so you would not hear a stuck solenoid if the switch jammed in no man's land!
So if the discussion turns to the jammed switch scenario, in addition to the original time-out method, you could install a 25-cent PTC resettable thermal fuse that would cut power to the coil if active for more than a few seconds. Of course the train might still derail if the switch is stuck in no man's land but at least the switch coil would not burn up! 
And, then, to peel even more layers of the onion, I suppose one could concoct a scheme to cut track power if the switch is stuck. Then the train would stop dead in its track, not derail at the switch, user gets some kind of fault indication, user clear the fault, resets the electronics and away we go...
