Here is an example code, of where the issue is coming from:

    use LWP::UserAgent;
    use HTTP::Request::Common qw(GET);

    my $ua = LWP::UserAgent->new;

    # Define user agent type
    $ua->agent('Mozilla/8.0');

    # Request object
    my $req = GET 'http://www.gazeta.ru/culture/2016/04/22/a_8191769.s
+html';

    # Make the request
    my $res = $ua->request($req);

    binmode STDOUT, ":utf8";
    print "Content-Type: text/html; charset=utf-8 \n\n";

    use Encode;

    if ($res->is_success) {

        my $title;
        $res->decoded_content =~ /<title>(.+?)<\/title>/ and $title = 
+$1;
    
       # prints correctly here!
        print "GOT TITLE: $title \n";
   
        $DB->table("ReadingGrabCache")->add( { title => $title, url =>
+ "Foo" });

        my $test = $DB->table("ReadingGrabCache")->select ( { url => "
+Foo" })->fetchrow_hashref;

       # buggered content here
        print "BLA: $test->{title} \n<br>";

    } else {
        print $res->status_line . "\n";
    }
[download]

The DB module is encoding insensative (i.e its not doing any kind of conversion) ... so I'm confused how it can be fine here, and then broken when grabbed back :(

Checking it in phpmyAdmin also shows the issue:

Румыния не будет участвовать в «Евровидении-2016» из-за денег - Газета.Ru

The table is quite simple... but maybe I've missed something:

CREATE TABLE IF NOT EXISTS `ReadingGrabCache` (
  `grab_id` int(11) NOT NULL AUTO_INCREMENT,
  `url` varchar(255) CHARACTER SET latin1 NOT NULL,
  `images` text CHARACTER SET latin1 NOT NULL,
  `title` text COLLATE utf8_bin NOT NULL,
  `description` text COLLATE utf8_bin NOT NULL,
  `all_images` longtext CHARACTER SET latin1,
  PRIMARY KEY (`grab_id`)
) ENGINE=MyISAM  DEFAULT CHARSET=utf8 COLLATE=utf8_bin AUTO_INCREMENT=
+141 ;
[download]