Since there are backslashes involved more care is required to get good results. However, this can be implemented with something like:

#!/usr/bin/perl
use warnings;
use strict;

my $regex_1 = '\\\\'; # Need pairs because of escaping
my $regex_2 = '__BACKSLASH__';
for (<DATA>) {
    s/\Q$regex_1\E|\Q$regex_2\E/\\/g;
    print;
}

__DATA__
8\\.
8__BACKSLASH__.
[download]

Of particular note, I've used \Q and \E to escape regular expression metacharacters - omitting this is my guess on why you were having issues. See Quote and Quote like Operators in perlop for info on these.

I was actually doing the exact same thing. The issue was I was reading the data through perl -d and was using the 'x' command. Should have used the 'p' command to print the data instead. That's why I was seeing that extra '\'. Thanks for your help though

for (
   '8\\\\.',            # 8\\.
   '8__BACKSLASH__.',
) {
    my $s = $_;
    $s =~ s/\\\\|__BACKSLASH__/\\/g;
    print "$s\n";
}
[download]

8\.
8\.
[download]

$string = '8\.' if $string eq '8\\.' or $string eq '8__BACKSLASH__.',
[download]