In general, forking and multi-threading won't speed up non-blocking algorithms unless you have a multiprocessor system and your program can take advantage of multiple processors. This doesn't look like one of those occasions.

Not being a Windows programmer, I merely suspect you could hoist the OLE constructor out of the loop for a minor speed boost:

my $fso = Win32::OLE->new( 'Scripting.FileSystemObject'  );

for my $Path (@ARGV)
{
    my $Match = {};
);
[download]

You'll also avoid one-character buffer flushes if you re-enable buffering on STDOUT. Just about everything else looks like a micro-optimization to me, and I can't benchmark it. Insert hand-waving here.