my @refs;
for (1..10) {
   push @refs, \my $foo;
}
print "$_\n" for @refs;
[download]

SCALAR(0x18317d8)
SCALAR(0x23605c)
SCALAR(0x236098)
SCALAR(0x236d88)
SCALAR(0x1831838)
SCALAR(0x1831808)
SCALAR(0x1831868)
SCALAR(0x1831880)
SCALAR(0x1831898)
SCALAR(0x18318b0)
[download]

You'll might see \do { my $foo } and do { \my $foo }, but the extra scope is almost always overkill.

I think that saying “the extra scope is almost always overkill” is a little disingenuous—your second example only works¹ because of the extra scope. Specifically, if you tried to unroll this for loop, then you'd have to either think up a bunch of extra names, or get warned.

¹ By which I mean, only works silently under warnings.
UPDATE: Despite typing it out myself, I didn't register the ‘almost always’, and so was reacting as if the statement was unqualified. I'm sorry about that.

your second example only works1 because of the second scope

What do you call the second example? The way I see it, I only posted one.

Specifically, if you tried to unroll this for loop, then you'd have to either think up a bunch of extra names, or get warned.

But why would you do that. I can't imagine any scenario where you'd do that. Can you come up with a real world example where you need to generate more than one such reference in the same scope? If there was such an case, it seems to me that having different names for different variables would be a good thing.