I have designed and tested a circuit which is intended to power a relay approx 1 second after power is applied and to turn off the relay as the power begins to drop.
This was designed because I have some DC-DC converter boards which give nasty large spikes of 4V - 10V of 100's of mS on the output when first powered on and when powered off and the input voltage to the DC-DC converter drops to 4V-2V.
i.e. ATX PSU AC-DC source - 12V output - DC-DC buck\boost converter ---- (relay contacts) -- output terminals.
^ - same 12V also powers delay circuit+relay
I designed the on-delay circuit in LTSpice and built it on breadboard.

It works exactly as expected. I use a 5V relay so that when power is switched off and decays, the relay will quickly turn off before I get the spikes from the DC-DC board which happens when the power rail gets to around 4V-2V. The Schottky diode is there to discharge the capacitor quickly so that if the power is turned on again quickly, the full time delay will occur again. Q2 and the Zener get a bit warm but not much above 60 deg C.
Now it occurred to me that I could add a SPST On/Off switch across the capacitor to provide a manual method of turning off the relay without needing to switch off the whole circuit.
This seems to work Ok at exactly 12V or below, but if I increase the source voltage slightly, shorting out C1 does not turn off the relay.
Once the 5V relay is on and latched, when I short out the capacitor the relay does not turn off. The relay has approx 1V across it which is enough to keep it on. The Q1e to Q2b connection also has approx 0.6-0.8V.
The relay is
http://www.farnell.com/datasheets/1870236.pdf but as far as I can see, the problem is that the transistors are still slightly on.
Any help appreciated.