Here's why I'm so stupid......
I've been working a lot lately with 18F series chips.
These are great because they have loads of different "fuses" which let you set up the chip in a particular way. One of the fuses on the 18F range allows you to reclaim pin one - the MCLR/PGV (programming voltage) pin and set it to an input. I do this with pretty much all my 18F chips.
So when I came back to the old 16F series chips (the 16F877A is an old favourite that first got me interested in PIC programming) I completely forgot that you need to apply 5v to the MCLR pin to stop the chip from resetting (pulling this pin low causes the PIC to reset completely).
So during testing, while trying to read the serial message(s) being "broadcast" from my 16F chip, I kept getting blackouts - periods when the serial stream would stop. Wiggle a few wires around on the breadboard and it would come back to life. For a little while.
Everything was pointing at a hardware problem - either a dodgy wire or a loose crystal or something. I re-wrote the firmware on both the slave and master chips, stripping each down to just the bare essentials, and used a really handy tool in the PICKit2 software to monitor the serial stream
You can connect anything that uses serial comms, and set the baud rate to any of the common bps values - 1200, 2400, 4800, 9600, 19200, 38400 etc as well as your own custom baud rate (ideal for debugging MIDI devices at 31250)
Anyway, after connecting everything up, setting it running and not touching a thing, the 16F would periodically shut down. I stripped out all the wires, rewired the board with a brand new set of jumper wires, put LEDs on all kinds of output pins and so on. I checked and double-checked the fuses, made sure the watchdog timer was disabled (and even tried enabling it and littering the firmware with watchdog reset commands) and checked the brown-out option wasn't selected. I even put a few different capacitors on the power rail of the 16F, just in case the input voltage was getting a bit wobbly.
The whole thing was suddenly a lot more complicated than any other PIC project I'd ever worked on and still it kept resetting.....
.... that's when I remembered that pin one (MCLR) has to be pulled high to stop the, erm, PIC from resetting all the time. Bugger.
It was as simple as that!
A single wire from pin one to 5V and everything worked beautifully.
If I had remembered about the MCLR pin earlier, instead of assuming it was always set to behave as an in input pin I think I'd have saved myself about two hours this evening.
Hopefully this post will help others in future to avoid wasting a similar amount of time on something so trivial!