There are use cases that the flakiness is in random testcases instead of a specific one. Such cases were seen in GPU tests (telemetry-based), and webkit_tests ( http://crbug.com/649463 ).

We could add an option for Findit to look at the result of the entire step or shard. But this could cause false positives more easily than looking at an individual testcase, because of longstanding flakes.
 

There are use cases that the flakiness is in random testcases instead of a specific one. Such cases were seen in GPU tests (telemetry-based), and webkit_tests ( http://crbug.com/649463 ).

We could add an option for Findit to look at the result of the entire step or shard. But this could cause false positives more easily than looking at an individual testcase, because of longstanding flakes.

Manual triggering seems better for such cases.
 

There are use cases that the flakiness is in random testcases instead of a specific one. Such cases were seen in GPU tests (telemetry-based), and webkit_tests ( http://crbug.com/649463 ).

We could add an option for Findit to look at the result of the entire step or shard. But this could cause false positives more easily than looking at an individual testcase, because of longstanding flakes.

Manual triggering seems better to start with for such cases.