Would using threads instead solve this problem?

Maybe. You could try something like this, which compiles clean but is obviously untested.

#! perl -slw
use strict;
use threads;
use threads::shared;
use Thread::Queue;

use constant NTHREADS => 20;

my @itemstoget = (); ## Get the items from somewhere?

my $Qwork = new Thread::Queue;
$Qwork->enqueue( join chr(0), @{ $_ } ) for @itemstoget;
$Qwork->enqueue( (undef) x NTHREADS );

my $Qresults = new Thread::Queue;

my $running : shared = 0;

threads->new( \&thread, $Qwork, $Qresults )->detach for 1 .. NTHREADS;

my $dbh = connectdb( 'blogdb' );

sleep 1 until $Qresults->pending;

while( $running or $Qresults->pending ) { ## Modified condition
    sleep( 1 ), next unless $Qresults->pending;

    my( $id, $url, $html ) = split chr(0), $Qresults->dequeue;
    if( $html ne 'FAILED' ) {
        # if the html meets the criteria for at least one client that 
+claims it,
        # extract the text

        if( checkhtml( $dbh, $id, $html ) ) {
            inserthtml( $dbh, $id, $html );
            print "OK  $id ",substr( $url, 0, 50 ),"\n";
        }
        else {
            print "SKIP     $url\n";
            dosql( $dbh,"update blogitems set getattempts=999 where id
+=$id" );
        }
    }
    else {
        print "FOREIGN or FAILED  $url\n";
        dosql($dbh,"update blogitems set getattempts=999 where id=$id"
+);
    }
}
$dbh->disconnect;

exit;

sub thread {
    { lock $running; ++$running }

    my( $Qwork, $Qresults ) = @_;

    my $user_agent = LWP::UserAgent->new;
    $user_agent->timeout( 30 );

    while( my $item = $Qwork->dequeue ) {
        my( $id, $url, $title, $excerpt )  = split chr(0), $item;

        my $request = HTTP::Request->new( 'GET', $url );
        my $response = $user_agent->request( $request );

        my $html = ( $response->is_success 
            and not isforeignlanguage( $response, $title, $excerpt, $u
+rl ) )
            ? resolve_charset( $response->content )
            : 'FAILED';

        $Qresults->enqueue( join chr(0), $id,  $url, $html );
    }
    undef $user_agent;
    { lock $running; --$running }
}
[download]

It should avoid the reentrancy problem with DBI by only accessing the DB from the main thread. It queues up the work items in a shared queue and starts 20 threads to fetch the urls. It then performs the sanity checks on the response before posting either html fetched or a failure code back to the main thread for processing to the database.

The threads will stop once the queue empties and the main thread will stop once there are no further results to process.