The multivm builder and perf bot is sometimes passing the webkit tests now.

There are a small number of tests that are flaky, and perhaps we should add them to the status file on our branch, so we stay green. On linux, a typical webkit tests result is:

Unexpected flakiness: text-only failures (3)
  http/tests/security/isolatedWorld/events.html [ Failure Pass ]
  http/tests/security/redirect-BLOCKED-to-localURL.html [ Failure Pass ]
  svg/dom/SVGScriptElement/script-set-href.svg [ Failure Pass ]

Unexpected flakiness: crashes (1)
  media/audio-delete-while-slider-thumb-clicked.html [ Crash Pass ]

Unexpected flakiness: timeouts (4)
  fast/workers/storage/interrupt-database.html [ Timeout Pass ]
  storage/indexeddb/key-generator.html [ Timeout Pass ]
  storage/indexeddb/odd-strings.html [ Timeout Pass ]
  storage/indexeddb/structured-clone.html [ Timeout Pass ]

Regressions: Unexpected text-only failures (1)
  compositing/iframes/scrolling-iframe.html [ Failure ]

So only the scrolling-iframe.html failure is turning the bot red. I'll also send a candidate CL to put this in the status file.

The windows bot is also almost green - only 2 timeouts turning it red. I'll put those in the CL. But macos has thousands of image mismatches - something needs to be fixed on the setup of the machine, I guess.

So do we want to keep running the webkit tests on these bots, and do we want the results to turn the bots red? There is also some flaky redness in sunspider and a couple other perf tests.