@enebo will have to answer with details, but I can say a few things.

This isn't really part of the arg list (or the signature of the block) at all. I believe it's handled in the parser by just forcibly declaring that variable name at the current level (depth 0), so it is from then on handled as local to the block. The fact that it was declared such is not directly reflected in any of the final parse artifacts. If it overlaps with an enclosing scope, it will just show up in the StaticScope hierarchy as an overlapping variable declaration. If it doesn't overlap, it may not be possible to tell a variable was declared this way.

@eregon I don't think it was intentionally removed but it maintains a a field and optionally a whole arraylist per argsnode. Memory is a huge problem for us so I guess if you really need to be able to determine this style of variable (we do not as we depend on StaticScope for this) we could maybe make a specialized ArgsNode so we only pay any extra memory cost when this feature is actually used.

@enebo I'm not quite following.
How can StaticScope be used properly to detect such variables?
Adding all variables from variableNames in the block scope seems to break tons of specs, and guessing the index only works in a debugger :p
Could we maybe add a method String[] getBlockLocalVariables() in StaticScope, is that possible?

One could add the variable when noticing a DAsgnNode with depth==0, but that seems a very indirect way too.

@enebo I ended up using the StaticScope, so feel free to remove blockLocalVariables-related things in ArgsNode, since it's never used anyway 😉