> 2) HTTrack cannot just download only not yet downloaded 
> data. Instead, it tries to download part of the data 
> already downloaded and mirrored. It is another serious 
bug, 
> perhaps connected with the previous.

Okay, I could not yet reproduce the issue, but it is 
possible that the background queue cleaner is "dropping" 
entries without dumping them in the cache.
I have secured the system by using the "finalized" flag of 
the cache entry. In this way, no entry will be dropped 
without being flushed (the detetion of the entry will cause 
the flush in last ressort)

I will upload soon a -test2 release ; which should prevent 
any missing entries.

I also forced the cache to store all responses from the 
remote server, even errors such as 5xx ones, to avoid 
redownloading bogus files during rescan.

I'm not totally sure that this will be sifficient ; I will 
remake more tests during the next week ; but the cache 
should be secured anyway.