The patch that affects the for loop is trickier than it looks.

At the moment all the payload-extracting operations (if (x) |y|, while (x) |y|, for (x) |y|) alias the y location to the payload location with a simple trick: an "empty" (with no backing memory) variable named y is created and its allocation pointer is simply set to whatever the unwrapped/iteration address is.
Now this behavior is source of several nasty bugs (#2915) by itself, but the implementation is also flawed (#4361) as variables cannot reside at address zero!
I think the first step to properly solve this problem is to decide what the "right" behavior is and then either amend all the cases outlined above or implement some workaround to handle the allowzero case.

My vote is for making the unwrap operation perform a copy as it's more in-line with how const x = y behaves (and I see many other people thought the same). If the user wants to avoid the copy and/or have a mutable reference they can simply prepend a * to the variable name and reap all the benefits without having to stop and think too much about how those constructs behave.

I'll drop that patch from this series so that we can discuss this problem without holding back the other good stuff.

My vote is for making the unwrap operation perform a copy as it's more in-line with how const x = y behaves (and I see many other people thought the same). If the user wants to avoid the copy and/or have a mutable reference they can simply prepend a * to the variable name and reap all the benefits without having to stop and think too much about how those constructs behave.

I agree with this, this will be the path forward. Thanks