Unfortunately, apart from adding CPUs and RAM, there is not much you can do. Ubooquity was not designed to support heavy loads (the initial use case is a small home server or a NAS).

The only thing you can try (if you did not already) is to give Ubooquity more RAM to work with.
This is done with the -Xmx Java option.
If you use the script I provide, just uncomment this line (remove the "#") and change the value if needed:

#MEM_OPT=-Xmx512m

I think the problem you had was fixed by the recompression step, not the deletion of the .DS_Store file.

What field would be use for sorting ? Publication date ?

The page split is done using quite convoluted CSS properties combinations.
I am not satisfied with the way it is done, I'll very probably rewrite this part once I have found a more reliable way to split pages using Javascript and CSS (I'm thinking "canvas").

In the meantime you can try something with the column-count property.
Eg:

#group
{    
    -webkit-column-count: 3; /* Chrome, Safari, Opera */
    -moz-column-count: 3; /* Firefox */
    column-count: 3;
}

.rootlink {    
    font-family: "Oswald";
    font-size: 30px;  
    display : inline-block;
    padding: 5px;    
    background-color: #a6a6a6;
       border: #080808 solid 4px;
    color: #1a1a1a;
    width: 150px;    
    box-shadow: -5px 4px 11px 2px rgba(0, 0, 0, 0.75);
    font-weight: bold;
    -webkit-animation: fadein 1s; /* Safari, Chrome and Opera > 12.1 */
       -moz-animation: fadein 1s; /* Firefox < 16 */
        -ms-animation: fadein 1s; /* Internet Explorer */
         -o-animation: fadein 1s; /* Opera < 12.1 */
            animation: fadein 1s;
}

Yes, the URL you see are not directly related to the directory structure of your collection. Nothing's wrong here (the "185" is an internal ID Ubooquity assigned to this book during the scan).

This white an grey cover is the default one, displayed when Ubooquity was not able to find the cover in the book:


It should not happen on Epub files (if you have an example of non-working Epub, I'd be interested in seeing it).
Mobi files are another story: the format is not public (and not documented) so Ubooquity extracts all images it finds (using a very empirical process) and tries to guess which one is the cover. That's the best I can do for Mobi files.

As for the mixed covers (the problem solved by a refresh), I have it sometimes on my iPad, but I have never been able to understand why. Never had it on Windows (with Firefox or Chrome).
Do you have this problem only with Chrome or also with Safari ?
(I don't own a Mac, so investigating this one will be difficult)