Handling of action retries using "failure_retries" isn't exactly correct
Here's a job log: https://validation.linaro.org/scheduler/job/2412808. The job has
failure_retry: 3 for the "boot" (effectively flash) action, and as can be seen in the log, 2 first attempts to flash failed, but 3rd finally worked. However, in test results we see 3 failures: https://validation.linaro.org/results/2412808/lava .
That gets also pulled into SQUAD, which reports overall failure for this action: https://qa-reports.linaro.org/lite/zephyr-micropython/build/d1087d2e-zd18ab4af-1546/testrun/4281193/suite/lava/tests/ .