Ugh. What if you have a title of 'Why "foo", "bar", and "baz"?' and it gets written to a CSV file as: ...,16,"Why "foo", "bar", and "baz"?",20,... then how do you expect to be able to tell which "s need to be escaped??

Well, I'll try me best... Let's assume that no title contains a string matching /",\S/ and that there is never whitespace after a comma in your CSV file.

s{
    \G(
        [^",]+
    |
        "(.*?)"
    )(,(?=\S)|$)
}{
    if(  ! $2  ) {
        $1.$3;
    } else {
        my $f= $2;
        $f =~ s/"/""/g;
        '"'.$f.'"'.$2;
    }
}gx;
[download]

If you do have whitespace after commas, then an alternate solution would be to assume that all titles that contain "s always contain an even number of quotes and that the first character after the first quote of a pair isn't a comma:

s{
    \G(
        [^",]+
    |
        "((?:
            [^"]+
        |   ""
        |   "[^",][^"]*"
        ))*"
    )(,|$)
}{
    if(  ! $2  ) {
        $1.$3;
    } else {
        my $f= $2;
        $f =~ s/"/""/g;
        '"'.$f.'"'.$2;
    }
}gx;
[download]

Unfortunatly, neither situation applies. Some of the titles have only one quote in them, some have three or four. Some of the quotes have commas after them. The solution I've decided to go with is editing the paradox database to get rid of the commas. Find and Replace, yeah, baby, yeah. Groovy.

I think it'll be the only reliable way to do it, and it'll probably fix some of the errors the database has been having.

/me crosses his fingers.