Not sure if this has been discussed, but I wanted to make the general remark about backwards-compatibility of introducing multi-byte opcodes. Introducing new multi-byte opcodes will influence jumpdest analysis and thus can have effects on all code following the point where the new multi-byte opcode is used in old code, not only the point itself:
If the argument to a newly-introduced multi-byte opcode is the byte value of JUMPDEST, then this was a valid jump destination before the hard fork, but it is no longer a valid jump destination after the hard fork.
Even worse, if the argument is a push opcode, the push data will be interpreted as code, while it was push data before. In the worst case, this might cascade all the way to the end of the code if some push data is the byte-value of jumpdest or of the newly introduced multi-byte opcode.
Because of that, I would advise to use account versioning for any change that introduces new multi-byte opcodes.