If I've understood you right, this should do it:

my %h;

# build a giant hash of all the info. Keys are ids, values
# are hashrefs whose keys are the source filename and whose
# values are the lines themselves.

while (<>) {
  my @fields = split ',';
  $h{$fields[0]}{$ARGV} = $_;
}

# for each id (lexically sorted)

for my $id (sort keys %h) {
  my @keys = keys %{$h{$id}};

# if it was present in only one file, print it and move on
  if (scalar @keys == 1) {
    print $h{$id}{$keys[0]};
    next;
  }
# if it was present in more than one, find out whether
# all the lines are the same by building a hash with
# each line as the key, then testing whether you end
# up with more than one key.

  my %cmp;
  $cmp{$_} = '' for values %{$h{$id}};
  print keys %cmp if scalar keys %cmp == 1;
}
[download]

Updated: Now I feel silly. This can be much simpler.

while (<>) {
  my @fields = split ',';
  $h{$fields[0]}{$_} = '';
}

for my $id (sort keys %h) {
    print keys %{$h{$id}} if scalar keys %{$h{$id}} == 1;
}
[download]

and, if one really wanted, the for loop could even be the gratuitously uber-terse:

scalar keys %{$h{$_}} == 1 and print keys %{$h{$_}} for sort keys %h;
[download]

I love Perl.

Updated again: You know how it goes. You start thinking about how something can be terser, and next thing you know, you're golfing.

perl -ane '$h{$F[0]}{$_}=0;END{keys%{$h{$_}}==1&&print keys%{$h{$_}}fo
+r sort keys%h}' f1.txt f2.txt
[download]

OK. I stop procrastinating now.