Setting maximum number of links scanned to 30 000 000 (30 million) generates a
nice memory hole (capable of eating all free memory). It also might me
something to do with fact that I am running two instances of winhttrack
together, but previously it worked fine(with maximum links set lower). When
scan starts, it shows under "Links scanned:" 0/-1  . Only one link is given to
process and is correctly written and page exists.

Sadly there are some pages that have so many links(dictionaries)