Just thinking about it - I don't think it's necessary to have two pins controlling VREF.

In the original circuit, there are three possible states:

GP 2is open circuit; VREF would range from 0 to ~1V (depending on R23)
GP2 is 0V; VREF would be 0
GP2 is 5V; VREF would be between 0 and 5V (again, depending on R23)

The last mode of operation isn't what I suspect is used, so I'll discount that.

However, GP2 can easily create a PWM signal by going from open-circuit to 0v, which at a 25% duty cycle would bring VREF from between ~0 and 0.25V. C23 is used to smooth the signal.

I'm beginning to have an idea about a (possibly) better way to do this.

GP2 is used for the CLOCK signal for the L297 - this is the PIC's interrupt pin, and would be easier to deal with there.

GP1 has a 10K trim resistor which is used to control the minimum hold voltage.

GP0 is connected DIRECTLY to VREF. This is used to monitor VREF.

GP5 is used to control VREF using the same circuit as the original does.

GP4 has a 10K trim to control the delay time for the circuit.

The advantages with this are:

1) The hold VREF can be easily adjusted, and measured with a volt meter
2) The PIC has voltage comparison circuits on GP0 and GP1, and this would be using them
3) Interrupts can easily be used for everything (VREF going below voltage, clock signal coming in, timeout occurring)
4) The PIC shouldn't need to be reprogrammed at all (once it's all working)

If there was a spare output pin, then the L297's RESET pin could be controlled (removing the need for R15, D12 and C16 on your circuit). If we didn't monitor the VREF signal (GP0), then this could be done instead - the trim for GP1 would actually control the minimum PWM duty cycle.