I don't think it would change the difficulty of adding new characters for keywords and operators. It would be no harder than adding @ to the basic source character set.
+1.
It is currently assumed, but not stated that the characters that are in the basic character set can be used in arbitrary grammar elements.
This doesn't have to be the case.
For example adding "@" to the basic character set would not mean we could use it as an operator.
Nor apparently that it would have to be representable in the physical character set. But it would have to be representable in the execution character set...
(At least what I propose would resolve that last point.)