In addition to the changes suggested by AnonyMonk above, I find it is wise to capture regex 'capture' variables and get them nailed down as soon as possible. This is because the evaluation of any further regex will probably cause the contents of these variables to... well, vary. (You say, "But I'm not using any other regex and never will." Famous Last Words. Nail them down. And, in fact, you do have another regex, the one in the second, nested while-loop. If you were ever to add a statement after the second while-loop that made reference to the  $1 $2 $3 $n   capture variables expecting that they would retain the values they had after execution of the first regex, you would be unpleasantly surprised. Nail them down.)

Maybe something like this. (Since you don't seem to use  $1 anywhere, I leave it out of account.) (Note the  /g modifier on the  m//xmsg regexes.) (Untested)

my $pkt_file = ...;
my $dat_file = ...;
...
open (my $fh_pkt, '<', $pkt_file) or die "ERROR Opening $pkt_file (Err
+or = $!)\n";
...
while (defined(my $line_in = <$fh_pkt>)) {
    if (my ($pkt) = $line_in =~ m{ \w+ \s+ (\w+) }xmsg) {
        my $out_file = "${pkt}_tlmval.txt";
        open (my $fh_out, '>>', $out_file) or die "ERROR Opening '$out
+_file' (ERROR = $!)\n";
        ...
        open (my $fh_data, '<', $dat_file) or die "...";
        while (defined(my $line_in = <$fh_data>)) {
            if (my ($foo, $bar) = $line_in =~ m{ \Q$pkt\E \s+ (\w+) \s
++ = \s+ (\d+) }xmsg) {
                print $fh_out "check '$pkt' '$foo' vs '$bar'\n";
                }
            }
        }
    }
[download]

Note that the  $line_in lexical variable in the nested while-loop is guaranteed to be completely separate and isolated from the lexical of the same name in the outer loop. Even so, I think I would prefer to use different names for these two variables just for the sake of self-documentation and maintainability.