You are right. A triac can be used in the described circuit with the A1 terminal to ++V rail. It works specifically because the device is a four quadrant device and will trigger with current coming out of the gate terminal.

As a matter of fact I see now that this is actually the way the circuit is meant to work. I made a mistake in my quick evaluation. The TLV431 will be OFF with no cathode current when the input side resistor divider (variable resistor in series with fixed resistor to GND) is set such that the ref input is below 1.24 volts. This would keep the triac OFF as long as the A1 terminal was at the V+ rail. As the rail voltage rises (or the variable resistor is made to a lower ohms setting) eventually the TLV431 ref input will get up to 1.24 volts. This will cause the TLV431 to suddenly turn on and conduct cathode current to GND. This will cause the triac to turn on hard as long as the ++V rail is at least 1.24 volts plus the trigger threshold voltage for the triac.

So the conclusion is that this combination of the shunt reference and the triac is a clever way to make a crowbar circuit. The TLV431 provides plenty of gain to make the switch point of the triac be a well defined snap point and not a soft sloppy soft turn on. This actually has some advantages as I see it over the zener diode / SCR circuit shown here by someone else which has way less gain and will end up having a soft knee turn on.

Michael Karas