This is always possible, but two things suggest to me that this is not what's happening. First, the output to the LED occurs in the interrupt handler before it ever gets to any of the jnb RI,$ instructions. Second, the exact same code when applied to port 0 works OK, and does output to the LED, and does not get stuck on any of the infinite loops.

I have only very indirect ways of telling what the code is doing, and this is very frustrating. I have no ICE, or logic analyser, or scope, so I have to rely on changing the code and observing indirectly signs of what it's doing. I've no doubt that when this is eventually cracked, it will turn out to be that the code is doing something that I didn't expect. I just need to spot what this is!

Rowan