nesdev.org

On the Commodore side we speculate that different versions of the CPU might not have the same illegal opcodes. This is because they are dependent upon the exact layout of the ROM, traces and in some cases length of the traces and process used.

Has anybody seen or done testing of the full set against the NES CPU, R6502, GTE6502, 6507 etc

I recall some random source stating that a very wide breadth of NMOS 6502 variants have been tested, and for every instruction that is not $8B ("sax #immed"), $AB ("lax #immed"), nor some of the KIL instructions, everything else works identically ... although sometimes inscrutably.