As far as I can see, the Metek at 30.tse10 never reported good w variances. spd, dir, w, u'u', and v'v' all look okay, but w'w' is always bad. It appears that, fairly regularly, the reported (parsed?) w from this sonic has a bad sample with a large + or - value. Should be fixable using min/max ranges in the .xml. Might want to look at the message though, to see if this could possibly be a parsing error, though this is unlikely since all the other Meteks were just fine.