If we ignore the relay for a moment, then the junction between R1 and VR1 would scale with the supply voltage. Current does not flow through ZD1 until Vr1 > ~ 11.1V. Then Vbe raises to about 0.7V at which point the Relay activates.
Then it stays activated. And thats where I am confused.
If R3 took its supply from the input then that would work, but then the relay would de-activate when you reduced the supply voltage - and it doesn't, TR1 stays hard on until the relay lets go at 3v. If you then raise again it activates again at about 7V demonstrating that the transistor has never switched off. So its latched.
Now, to get further into it I could dismatle it further, but since it now works and all I have to do is replace C1, I'm probably not going to risk breaking it. So it may end up keeping its secrets.
I have thought about this a bit more, T1 Must be an SCR. I don't come across those much these days and all the more important that the capacitors have some capacitance because otherwise it could trigger on spikes and will by definition latch on. I'll see if I can find markings before I stick it all back together. I could also easily test this by desoldering the "base" and triggering it manually, if it stays on then thats what it is.
So whilst I know what an SCR is and how its triggered, what I don't know is what voltage will be present on the gate once its triggered (if its not connected). I mean that we know you need a threshold to trigger it but we see 0.8V after the device latches on. Now, since its a stack of silicon that may well be correct. Hopefully someone understands what I meant there.
I had a look at whats available today and T092 case style Thyristors with a gate voltage of 800mV do exist and perhaps the part number is irrelevant from the 60's anyway.