I have thought about it but not implemented it. The one problem I can see with it is that instead of using the inbuilt UART you would have to make a software one as you would need to be able to have direct access to the data line the DMX signal comes in on from Rs485 transceiver. You could buffer a copy of this signal to another input pin and use interrupts.

Just out of interest how unreliable is your current method of detection? Is there a noticeable delay in receiving frames that you can see or are you using test equipment to tell if you are missing frames? Because the Break has to be at least 88uS long and a chracter is only 44uS long it should nearly always pickup the break as a frame error shouldn't it. In my code after getting the Frame error I check that the first valid character is equal to 0x00 , the start.

Getting back to the interrupt idea you could set it that a low on the input pin starts a timer running which is set for 88uS, the length of the break, every time the input pin goes high this timer gets reset. If it makes it to 88uS then there is a break. (I think some early DMX512 gear used counters and timers for this before micros became popular.)

I am not sure if this was the discussion you were looking for but the following link talked about a similar issue. In this case how to make sure the MAB was valid:

The link is:
http://members3.boardhost.com/dmx512/msg/2689.html

Please don't take anything as I have said as a critisicm of your idea. I have set up a PIC to control some relays from DMX512 so I am very interested in looking at this reliability issue.