http://sn.1w6.org/file/mk-20140626T074125-s4nljpz.html
@mk if I understand the output correctly, NO bad blocks have actually been recorded because the 'multiply-claimed blocks' are shared with "the bad blocks inode": apparently it won't write those bad blocks before cloning those blocks first. OK. (but why did that happen? should "bad blocks inode" not be protected from being overwritten by other files?)