I started looking at this based on the data as presented; however, there appears to be a question mark over whether all that whitespace actually exists. Here's solutions for both scenarios.

Without extra whitespace:

#!/usr/bin/env perl

use 5.010;
use strict;
use warnings;

my ($host, %seen);

while (<DATA>) {
    chomp;

    if (m{^hostname\d\.com}) {
        say $host = $_;
    }   
    else {
        say unless $seen{$host}{$_}++;
    }   
}   

__DATA__
hostname1.com
Gateway FAIL
Gateway FAIL
Gateway FAIL
hostname2.com
Gateway FAIL
Gateway FAIL
Gateway FAIL
[download]

Output:

$ pm_del_group_dup_nospace.pl
hostname1.com
Gateway FAIL
hostname2.com
Gateway FAIL
[download]

With whitespace (as written):

#!/usr/bin/env perl

use strict;
use warnings;

my ($host, %seen);
my $host_re = qr{ \A ( hostname \d+ [.] com ) \s* \z }msx;
my $fail_re = qr{ \A ( .*? ) \s*? \z }msx;

{
    local $/ = "";

    while (<DATA>) {
        if (m{$host_re}) {
            $host = $1;
            print;
        }
        else {
            print unless $seen{$host}{(m{$fail_re})[0]}++;
        }
    }
}

__DATA__
hostname1.com

Gateway FAIL

Gateway FAIL

Gateway FAIL

hostname2.com

Gateway FAIL

Gateway FAIL

Gateway FAIL
[download]

Output:

$ pm_del_group_dup.pl
hostname1.com

Gateway FAIL

hostname2.com

Gateway FAIL
[download]