in reply to Re^7: ... for (@_) x= 2; (scalar assignment)
in thread ... for (@_) x= 2;

Ah, ok.

I know three cases where parens matter (aside from when they override precedence or disambiguate syntax):

I think there's another one.

For the first two, it was either that or create a new operator. No other operators behave differently based on whether one their operand is list-like or not. It's a pity = and x didn't adopt the same definition of list-like, but then again, one expects lvalues and the other expects rvalues.