Given that the exploit relies on using byte sequences that cannot be interpreted as valid utf8 strings, I would think that anyone writing a script that uses the "-T" flag, and expects to handle utf8 data from a tainted source, would prefer to read such input as ":raw", and always use Encode::decode() to convert it to perl-internal utf8 form.
And in doing so, it would usually be prudent to do it like this (adapting the sample code given in the OP):
#!/usr/bin/perl -T
use strict;
use Encode;
%ENV = (
PATH => '/usr/bin'
);
open my $filehandle, "< :raw", "test.bin" or die $!;
my $word = readline $filehandle;
eval { $word = decode( "utf8", $word, Encode::FB_CROAK ) };
if ( $@ ) {
warn "unusable input from test.bin\n";
}
else {
my ($untainted) = $word =~ /^(\w+)$/;
if ($untainted) {
# It passed the regex, so it is "safe".
system "echo $untainted";
}
}