> Does Httrack have in fact a limit of links to be scanned 
or 
> a relation memory<->links?
The reasonnable limit is few millions links. But the more 
you allow to crawl, the more memory httrack will eat.. for 
large scale mirrors (such as backuping multiple sites for 
preservation), you'll need more memory, and possibily 
experience slowdowns (the internal hashtable will be 
overloaded a bit)

> I've do several attempts but after 2 or 3 days of machine 
> work the program dies peacefully without errors, without 
> apparent work (that is nobody is reading the HD).

No report in hts-log.txt ? The memory was exhausted, maybe ?