though if I need to go to the bother of calling waitpid etc. I will probably just bite the bullet and use Win32::Process instead.

Why? Why is "the bother of calling waitpid" so onerous?

This just ran to completion in 10 seconds or so(Note:1000):

use strict;
use warnings;

$| = 1;
my $cmd = qq{$^X -le "print 'pid=', \$\$"};

for my $i ( 1 .. 1000 ) {
    print "$i: run $cmd ---------\n";
    my $rc = system( 1, $cmd );
    waitpid( $rc, 0 );
}
[download]