When bug 1473786 landed, it caused some mochitest bustage, which came with a screenshot that looks like it was taken before the first error happened: https://taskcluster-artifacts.net/AKmGTGMpRw-746We-6XbNg/0/public/test_info/mozilla-test-fail-screenshot_Ek2PHY.png And the logs showed this for the first test failure: 17:42:19 INFO - TEST-UNEXPECTED-FAIL | dom/canvas/test/test_imagebitmap_extensions.html | [YUV422P -> RGBA32] pixel(1,0)channnel(R): dstImageDataValue:255 should equal to dstDataValue_R: 28 17:42:19 INFO - window.onmessage@dom/canvas/test/test_imagebitmap_extensions.html:31:9 17:42:19 INFO - EventHandlerNonNull*@dom/canvas/test/test_imagebitmap_extensions.html:29:5 17:42:19 INFO - Not taking screenshot here: see the one that was previously logged So, that first screenshot, that was taken for $reason, is preventing an actual screenshot from being taken of the actual failure. Full log: https://treeherder.mozilla.org/logviewer.html#?job_id=188481017&repo=autoland&lineNumber=7416 Bug 1478269 is on file for the test failure itself.