Re: whether to allow x, y, *z on the RHS, I guess PEP 526 again doesn't specify this either way. I vaguely recall that during the discussion we weren't too sure about various extensions, but perhaps the most important thing we didn't want to allow was multiple variables (e.g. x, y: int would be ambiguous).

I'm okay with allowing that, and it's probably better to do it sooner than later. The fix presumably is very simple (just update the grammar) but I'd request a test to make sure we didn't miss anything obvious in the code generation. (But x: Tuple[int, ...] = (1, 2, *xs) does work, so probably nothing.)