Ok, I have run a difference between the two sets of EDIT ROMS:
CBM8032_F32_103.bin
edit-4-80-b-50Hz.901474-04.bin
In the former case, is it true that the ROM is 4K?
A 'normal' EDIT ROM is 2K. Yes, I can see some changes above the 2K limit...
I have run a difference and the two sets of ROMs are very similar:
I am going to compare the ROM differences to the disassembled source code for BASIC 4:
http://www.zimmers.net/anonftp/pub/cbm/src/pet/pet_rom4_disassembly.txt.
The first thing that I notice is that the initialisation entry point at $E000 has changed from JMP $E04B to JMP $E9A0. This alone is outside of the E000...E7FF limit (2K) so puts the execution outside of a standard EDIT ROM. Of course, this may not be a standard EDIT ROM!
There does appear to be a block of code from $E900 to $E9Fx, so there is something there. Perhaps a bit of disassembly here required?
The first major block of changes is at $E6D0 - and this just happens to be the keyboard decoding table! I note there is a slight difference in the initialisation of the CRTC also.
The remaining changes are 'interesting' as they occur outside the 2K limit. This implies that some of the EDIT ROM (assuming it is really a 4K device) is enabled overlapping the I/O space. Presumably 'unprogrammed' ROM bytes ($FF) do not interfere with the normal operation of the I/O space due to the 'wire-OR' nature of TTL signals.
Just thought I would post this preliminary analysis...
Dave