I would characterize the effect of the code as:

For each input line, process it and extract an id. When run with no arguments, count in %Y the number of times each id appears; when run with a list of ids as arguments, count only how often the specified ids appear (with multiplicity).

I suspect you have been misled by a misinterpretation of:

foreach (@ARGV) { ++$Y{$y}, next if $y eq $_; }
[download]

which is equivalent to:

foreach (@ARGV) {
    if ($y eq $_) {
        ++$Y{$y};
        next;
    }
}
[download]

(I assume the useless next is left over from cutting down the original code.)

This is an idiom I actually use quite commonly for compactness when parsing arguments:

while (@ARGV && $ARGV[0] =~ /^-/) {
    my $arg = shift @ARGV;
    last if $arg eq '--';
    ($max = $arg || shift(@ARGV)), next if $arg =~ s{^-x}{};
    $inject = 1, next if $arg eq '-i';
    ++$debug, next if $arg eq '-d';
    die "Unknown option '$arg'\n";
}
[download]