This approach is unreliable. On my machine the script:

use Data::Dumper;
print Dumper({
  1007 => "Hello",
  1195 => "World",
});

print Dumper({
  1195 => "World",
  1007 => "Hello",
});
[download]

produces the output:

$VAR1 = {
          1195 => 'World',
          1007 => 'Hello'
        };
$VAR1 = {
          1007 => 'Hello',
          1195 => 'World'
        };
[download]

The issue is that Data::Dumper just returns keys in the order that keys gave it, which returns keys that are in the same bucket in the insertion order. With a 2 key hash your odds of hitting this are 1/8. But when the hashing algorithm works right, a fixed portion of the keys wind up in a bucket with a neighbour. Therefore for larger hashes you are virtually guaranteed that the set of keys alone does not determine what order they come back in.

If this explanation confuses, then Re (tilly) 4: Flip Flop III - Musical Buckets may help.