Where is this fork you keep mentioning? Are you doing if (!fork()) { system(...) } or if (!fork()) { exec(...) }? If you do the former, you'll have two copies of the parent perl script running once the child exits.

Example use of launching an executable using fork():

# We are in the only process.
printf STDERR ("pid = %d\n", $$) if $DEBUG;

$pid = fork();
die("fork failed!\n") unless (defined($pid));
if (!fork())
{
   # We are in the child process.
   printf STDERR ("child pid = %d\n", $$) if $DEBUG;

   exec(...);
   # We'll never reach here if exec() worked.

   # Avoiding die(); we don't want eval{} catching this.
   print STDERR "exec failed!\n";
   exit(2);
}

# We are in the parent.
printf STDERR ("parent (pid = %d) successfully spawned child (pid = %d
+).\n", $$, $pid) if $DEBUG;
[download]

Alternatively:

use Win32;
use Win32::Process;

sub GetLastErrorStr {
   return Win32::FormatMessage(Win32::GetLastError());
}

my $process_obj;
Win32::Process::Create(
   $process_obj,
   "bla.exe",
   "arg1 arg2",
   0,
   NORMAL_PRIORITY_CLASS,
   "."
) || die("Error running bla.exe: " . GetLastErrorStr() . "\n");
       
printf STDERR ("parent (pid = %d) successfully spawned child (pid = %d
+).\n", $$, $process_obj->GetProcessID()) if $DEBUG;

...

$process_obj->kill(255);
[download]

By the way, in Windows, I don't think the process has a chance to shut itself down properly when told to die using kill(). If the application has a window, it's better to send WM_QUIT to the window (and kill it if that doesn't work). If it doesn't have a window, it provides some kind of API (usually a DLL call) that will shut it down. This could explain your 100% usage after killing the process.