>>The problem I've run into is the oscillator. While the programmer is doing
>>its thing, it is supplying +5 to the F84, and the rest of the circuit,
>>which makes the oscillator do its thing, and thus the programming fails.
>
>Have a loop-back connector that grounds the oscillator (OSC1 pin) when the
>programming jack is connected.


This is interesting. Our F84 system does not halt the clock, and so far we
have a 100% success rate.  8 mhz clock, FWIW

We did hit a snag early on, in that the cable used in normal operation plugs
into the same connector, and serial input (soft uart) was causing the uP
clock to halt for 200-300uS.  This was traced to capacitive coupling in the
cable, combined with the diode on MCLR taking the MCLR voltage above 5V.

Our current "operating" cables have that conductor missing :)

I would have done it differently, but the modular connector is already the
largest component on the board, and another connector was NOT an option.