The ultimate HTTP_REFERER solution for PHP back button?

Question

Yes I know, another HTTP_REFERER effort. I cringe when I see it. But it has been handed to me as a solution to a 'go back' problem and it actually works, so...

...the problem is, I don't understand it enough to feel comfortable deploying it live.

I've gone through many many messages here about the perils of HTTP_REFERER, however I am told this code addresses them for this particular task... but I'm not convinced.

So, my Questions are at the bottom of the code below - but FIRST, here is the basis behind it:

This code goes into a shopping cart. It provides the action of the 'Continue Shopping' button after a user clicks 'view cart'. Its goal is to provide the most suitable (or, expected) redirection back to where the user came from.

By default, without this code added, the 'Continue Shopping' button of this particular 'view cart' page simply takes the user to the home page of the cart - which sort of blows if the user has navigated a few pages deep to find a set of products or a specific category, or was looking at a product, etc. So we want to improve on that. Also, the cart requires javascript to complete a purchase, which the user is warned of if they visit with JS off, so JS on is expected of the user during this operation. Finally, we wanted to keep the solution to a single block of code, so it can be applied easily and transported to updated versions of the cart with little issue.

The methodology as it was described to me:

1. If a referer is set AND it is not empty AND yes it DOES contain this cart's domain then

   1a. If the referer is the cart page itself, apply a javascript (-1) effect otherwise we'll be stuck in one place

   1b. ELSE go ahead and apply the Referer - Done.

2. ELSE if a refer is set AND it is not empty BUT no it does NOT contain this cart's domain then

   2a. Something is amiss, send them to the home page. (?? why ??)

   2b. ELSE let's apply the javascript (-1) effect. - Done.

The code:

if ((isset($_SERVER['HTTP_REFERER']) && !empty($_SERVER['HTTP_REFERER']) && (strpos($_SERVER['HTTP_REFERER'], HTTP_SERVER) === 0))) {
    if ($_SERVER['HTTP_REFERER'] == HTTP_SERVER . 'index.php?r=cart') {
        $this->data['continue'] = 'javascript:history.go(-1)';
    }
    else {
        $this->data['continue'] = html_entity_decode($_SERVER['HTTP_REFERER']);
    }
}
else {
    if ((isset($_SERVER['HTTP_REFERER']) && !empty($_SERVER['HTTP_REFERER']) && (strpos($_SERVER['HTTP_REFERER'], HTTP_SERVER) !== 0))) {
        $this->data['continue'] = 'index.php?r=home';
    }
    else {
        $this->data['continue'] = 'javascript:history.go(-1)';
    }
}

My Questions:

1. Ok, so we've combined two suspect methods, and collectively it seems to work. But what can go wrong with this (Security / Function)? Is there a showstopper here that I'm missing?

2. What is the second part (2. and 2a.) going to accomplish? It is explained to me to be an effort to determine if someone is purposely trying to feed a false referer - but it doesn't make sense to me - why do we care to route that person to the home page vs. using the javascript (-1) method?

3. ANY thoughts on improvement? Those always come in handly ;-)

Thank you for your time in helping with this...