Thought you might like to see my refactoring of your algorithm.

Same performance, but somewhat clearer implementation:

hdbm => sub {
    my $r = shift;
    my $i = 1;
    until( substr( $$r, $i ) eq substr( $$r, 0, -$i ) ) {
        my $j = index( $$r, substr( $$r, 0, $i ), $i+1 );
        $i = $j > 0 ? $j : ++$i;
    }
    return substr( $$r, 0, $i )
},
[download]

Specifically, it avoids the unfounded "infinite loop" charge without needing to introduce a redundant test.