my %hash = ('a'=>1, 'b'=>1, 'c'=>1);
@hash{'a'..'c'} = 2;
[download]

This has the effect of assigning 2 to $hash{'c'}. ...the last element in the slice.

That would be extremely inconsistent indeed, as @hash{'a'..'c'} is supposed to be equivalent to ($hash{'a'}, $hash{'b'}, $hash{'c'}). And it is. What I get for \%hash, with Data::dumper, for both, is:

$VAR1 = {
          'a' => 2,
          'b' => undef,
          'c' => undef
        };
[download]

both with 5.6.1 and with 5.8.0 (though the order might vary). <sigh-of-relief/>. So it is consistent.

Code used for testing:

my %hash = ('a'=>1, 'b'=>1, 'c'=>1);
@hash{'a'..'c'} = 2;
use Data::Dumper;
print Dumper  \%hash;

%hash = ('a'=>1, 'b'=>1, 'c'=>1);
($hash{'a'}, $hash{'b'}, $hash{'c'}) = 2;
print Dumper  \%hash;
[download]