Okay, I have now observed and tested the workaround for about two weeks. I
noticed a few sites which won't deliver any data when the headers were altered
as mentioned. That's easy to detect since the mirror already fails on the
first link.

However, whereever it works, it works perfectly. So far I've mirrored a few
million pages from formerly failing sites without a single "accidental
binary".

Nevertheless, it would be beneficial if this bug could be fixed in HTTrack.

Best regards, Klaus