Why nobody is answering that the most normal thing to do should be redirecting via .htaccess the URL with the ending trailing hash to the one without (or viceversa)?

I mean, that's what I would do if it was just the case of the home page.

Said that, I would not worry that much, because that is one of those exact duplication cases that Google understands and solves with no problem.

Your mistake is that you probably used relative URLs and not absolute ones (these are the ones like: http://www.domain.com/)

So, the hreflang in sitemaps.xml for your home page should be this (please note that I put /it/ but you should put the real italian version home page, and that is valid for all URLs:

<url><loc>http://www.domain.com/</loc></url>

As for your second question - mixing matching and not matching URLs in the hreflang annotations - I would not suggest it.

Yes.

One fundamental question: is the redirection based on IP detection or user agent?

Because if it is over IP detection, then that's something Google deprecates.

Instead, if it based over User Agent, then 302 and 301 are substantially the same for Google in this case, even though with a 301 you are sure to pass link equity.

On the other hand, why have not you considered these two options:

1. Making the root domain luxury.com the country selector page, so people land there then choose the country they want;
2. Dedicating the luxury.com root to the main country target, not using redirections at all and adding a over alert to all the people entering in the root that exist a version targeting their country (alert based on IP Detection, this time correctly).

Sorry for viewing this just now... but - forgive me if I am wrong due to a bad understanding of the question - but I think Tom answer is not correct.

You are telling that your main site is in English, but that has also a Spanish subdomain with just half of it localized in Spanish.

If this is the correct interpretation of the origin of your doubts, than, in the Spanish subdomain the hreflang should be implemented so:

IN CASE OF SPANISH SUBDOMAIN URL WITH SPANISH CONTENT

<loc>http://www.example.com/</loc>

IN CASE OF SPANISH SUBDOMAIN URL WITH ENGLISH CONTENT

<loc>http://www.example.com/</loc>

Why? Because those "en" and "es" mean "English Language" and "Spanish Language", so you cannot declare as Spanish something that Spanish is not. As well you cannot declare both URLs as to shown to English speaking users, because that would create an hiccup to Google, who would not know what of the two it has to finally show to English speaking users.

More over, if you don't want to extend the use of the hreflang suggesting also the countries where to show some given URL, then you should canonicalize the spanish.domain.com URL with English content to the original www.domain.com URL.

The idea of using also the country code ISO could solve - somehow - this issue, because writing something like this:

<loc>http://www.example.com/</loc>

Then you will be telling Google to show the spanish.domain.com URL to the people using english in Spain (Google.es), and the English one to all the people speaking English in the rest of world.

Be aware, though, that Spanish people using Spanish will see in the www.domain.com URL in their Google.es SERPs, because the x-default is telling Google that all the people not using the language indicated in the hreflang="x-X" annotation (which is English), will have to see the main domain URL, and not the spanish subdomain one.

Hreflang is quite a sudoku, but it is extremely logic.

The answer is very simple, independently if you are publishing the post on subfolders or subdomains or even country code domains names:

1. The canonical URL of the copies must be the URL of the original article. This is must be so because the content of all the articles is the identical. If you don't use the rel="canonical" as I suggest you'll fall in the duplicated content issue;

2. You use the hreflang annotations to tell Google what post's URL to show to the users using English and searching from the different blogs' countries.

Remember: the hreflang is not meant for solving the duplicated content issues, but only for suggesting what URL to show to users in a given language and country.

For duplicated content issues the solution is the rel="canonical".

I'd add one condition:

we do really want our images to rank in Google Images? Maybe we don't want that, especially since Google Images is not anymore that great source of traffic it was before its layout changed (we can see photos directly inside Google without clicking the "see the page" link). If it is so, I would prioritizes the image sitemaps.

But if you want your images indexed, picked up (and hotlinked) by other site or directly stolen but with you being able to find who used them without credit, then the image sitemap is needed:

here an example directly by Google: http://www.google.com/schemas/sitemap-image/1.1/sitemap-image.xsd.

I don't know what CMS are you using, but if it WordPress, the WordPress SEO by Yoast plugin automatically generates the image sitemaps.