The grammar in the doc is not the one used to generate the
parser/compiler.  The former is meant to be easier for humans to read,
the latter easier for the parser generator.  Neither completely embody
Python's syntax rules.  Additional restrictions may be applied later in
the chain.  From:
http://svn.python.org/view/python/trunk/Grammar/Grammar?rev=65872&view=auto

expr_stmt: testlist (augassign (yield_expr|testlist) |
                     ('=' (yield_expr|testlist))*)

augassign: ('+=' | '-=' | '*=' | '/=' | '%=' | '&=' | '|=' | '^=' |
            '<<=' | '>>=' | '**=' | '//=')
# For normal assignments, additional restrictions enforced by the
interpreter

Perhaps that should say "For augmented assignments....

Except maybe for the comment above, I think this issue can be closed.