The following layout test is failing randomly on Linux.
Apart of being a very slow test, it is failing randomly in the Linux bot, together with a few other editing/selection slow tests:
editing/selection/caret-at-bidi-boundary.html has been passing consistently for a long time, until we enabled threaded compositor. I'm reassigning it to bug #160119. Leaving this open to track the remaining failures.
This test(s) has been consistenly passing in the last 4000 revisions. Closing bug.