(In reply to Tom de Vries from comment #10) > We did run the tests before, and it started failing only recently, so in > that sense it's a regression. Can we quantify "only recently" better? Maybe I can bisect the 15SP3 kernel and see what has caused this...