Let me quote:

And the tests are not correct BTW. Time required for creating new thread depends on the size of all existing variables.

Yes. That is an annoying detail of the ithreads implementation. But, it is quite easy to avoid; you just spawn your workers early, and have them require rather than use what they (individually) need.

So, if worker threads require some heavy module, you have to use it before starting these threads (and I don't mean cases where you need just single worker thread), and it may seriously increase worker threads startup time, and it is not easy to avoid.

Yes. And if every worker fork needs to modify 1 bit on every 4x page of 1 GB of shared data, it's gonna take forever.

You show me your use-case, and I'll show you mine.

And if every worker fork needs to modify 1 bit on every 4x page of 1 GB of shared data, it's gonna take forever.

Rubbish

use strict;
use warnings;
use 5.010;
use Time::HiRes qw(time);

my $data = 'x' x 500_000_000;
my $t = time;
$data =~ tr/x/y/;
say time - $t;
__END__
0.469923973083496
[download]

use strict;
use warnings;
use 5.010;
use Time::HiRes qw(time);

my $data = 'x' x 500_000_000;
my $t    = time;
unless (fork) {
    $data =~ tr/x/y/;
    exit;
}
wait;
say time - $t;
__END__
0.765344142913818
[download]

use strict;
use warnings;
use 5.010;
use threads;
use Time::HiRes qw(time);

my $data = 'x' x 500_000_000;
my $t    = time;
threads->create(
    sub {
        $data =~ tr/x/y/;
    }
)->join;
say time - $t;
__END__
1.29563307762146
[download]