So I think it's better to keep the $0/$n symmetry.

I think having a "message" object is a better idea [than $$'s one]

What I like with the $$ idea, is that it would provide a simple way to merge creation arguments with variable arguments, i.e compose a message with both the abstraction arguments and the incoming message elements.

I have to say I quite like the "$$" idea as well, assuming that we can take the risk of breaking a few patches (if any).

I don't think it's a good idea to add a new object just for this functionality. For me this would create unnecessary complexity (you have to learn yet another object).

I'm not sure either. To me, both $0 and $1 etc. can be used to identify an instance of an abstraction.
IMO $0 is the quick way, but has the limitation to make it (nearly) impossible to access members from the outside.
That's why it often happened to me to rename an instance [myAbs] to e.g [myAbs myabs1], then to replace $0 in [myAbs] with $1, so I can easily access [myAbs]'s members from the parent - from anywhere in fact (Actually, nowadays I tend to use as few $0 as possible).
If we could use $0 in messages, then the last operation would be more complicated (cause you couldn't simply substitute $0 with e.g $1).

I agree that if we get the "$$" syntax, then it makes more sense to use "$$0" for the $0 argument! Without the "$$" syntax, I wouldn't see the problem...

One downside of using "$$0" is that it wouldn't be compatible with Pd-L2Ork / PurrData.If they have already diverged significantly, we probably don't have to care, but otherwise...