With test verification runs, runs are broken down into 4 steps:

1. Run each test 10 times in one browser
2. Run each test 5 times in a new browser each time
3. Run each test 10 times in one browser, in chaos mode.
4. Run each test 5 times in a new browser each time, in chaos mode.

Because chaos mode can extend the test duration, it's possible that tests failing in step 3/4 may be benign with a matter of accounting for chaos mode timeouts. Unfortunately, these intermittent failures can stick around for a while, like bug 1698240. It'd be nice if reporting/parsing on treeherder is improved to call out what step it failed on, so that we can root cause earlier.

The test log contains a failure summary, like

https://treeherder.mozilla.org/logviewer?job_id=387079330&repo=autoland&lineNumber=3213-3224

[task 2022-08-11T21:49:12.130Z] 21:49:12     INFO - :::
[task 2022-08-11T21:49:12.130Z] 21:49:12     INFO - ::: Test verification summary for:
[task 2022-08-11T21:49:12.130Z] 21:49:12     INFO - :::
[task 2022-08-11T21:49:12.130Z] 21:49:12     INFO - ::: toolkit/components/printing/tests/browser_print_page_range.js
[task 2022-08-11T21:49:12.131Z] 21:49:12     INFO - :::
[task 2022-08-11T21:49:12.131Z] 21:49:12     INFO - ::: 1. Run each test 10 times in one browser. : Pass
[task 2022-08-11T21:49:12.131Z] 21:49:12     INFO - ::: 2. Run each test 5 times in a new browser each time. : Pass
[task 2022-08-11T21:49:12.132Z] 21:49:12     INFO - ::: 3. Run each test 10 times in one browser, in chaos mode. : FAIL
[task 2022-08-11T21:49:12.132Z] 21:49:12     INFO - ::: 4. Run each test 5 times in a new browser each time, in chaos mode. : not run / incomplete
[task 2022-08-11T21:49:12.132Z] 21:49:12     INFO - :::
[task 2022-08-11T21:49:12.133Z] 21:49:12     INFO - ::: Test verification FAILED!
[task 2022-08-11T21:49:12.133Z] 21:49:12     INFO - :::

That can be hard to find (easy to miss) in the treeherder logviewer.