What's really specific in this case?

"foo" is a name which has been explicitly given, whereas "arg0", "arg1" etc. are the default names returned by the runtime.

I think it'd make sense to prefer explicitly names (when given) over default ones to avoid the situation that a parameter name is specified on a sub-type but not on the super type and then the default name from the latter is applied. In case explicit names for both are given, the one from the super-type should win.