So it's a matter of merging hashes? You could use an iterator, perhaps something like the following:

#! /usr/bin/perl  -l

my $zoo = {
  Mountain => { Dragon => 'Deadly', Troll => 'Deadly', Nymph => 'Oread
+', Snake => 'Deadly' },
  Cave => { Dragon => 'Deadly', Spider => 'Poisonous', Troll => 'Deadl
+y', Ogre => 'Deadly' },
  Forest => { Elf => 'Dark', Troll => 'Deadly', Nymph => 'Alseid', Spi
+der => 'Hairy' },
  Lake => { Nymph => 'Naiad', Snake => 'Poisonous', Serpent => 'Deadly
+' },
  Sea => { Kraken => 'Monstrous', Serpent => 'Monstrous' }
};

sub each_uniq {
  my @refs = @_;
  my %seen = ();
  sub {
    while (@refs) {
      my @pair = each %{$refs[0]} or shift @refs and next;
      return @pair unless $seen{$pair[0]}++;
    }
    return;
  }
}

my $iter = each_uniq(values %$zoo);
print join(", ", $iter->()) for 1..5;
[download]

Update. I read the previous questions monkini has asked, and it would appear the problem is largely unrelated to questions asked. From what I gather, he has a number of clusters each with a number of (labelled) points, and distances between points given. The problem apparently is to list the pairs of points in a cluster sorted by their distance, for each cluster separately. If so, the solution is to map clusters to points (HoA), iterate, slice map slice sort the HoH, print.