Testing for interactions between Griffin-enabled features and existing features

[fmarier]

Many of our features land behind a flag that is disabled by default. This enables gradual rollout of features via Griffin. Those features will typically contain a browser test which enables the flag before testing the feature. Therefore the feature itself passes its own tests prior to merge and rollout.

There is however a gap in our testing: browser tests for other features are never run with that new flag enabled. Instead, they run with the default value of the flag ("off") and so the interaction between the new to-be-rolled-out-later feature and existing features is not tested prior to rollout via Griffin.

One way to close that gap would be to require a PR which turns the flag on by default and runs through all browser tests successfully in CI, prior to enabling the feature for any users via Griffin.

Additionally, some of our features require supporting files to be shipped via the component updater before doing anything. We should ensure that all components are up to date before running these browser tests.

I suspect that these gaps in our testing explain why the regression described in brave/brave-browser#32462 was not caught quickly despite the fact that we do have a browser test for the query string filter.

[fmarier]

Moved to https://github.com/brave/devops/issues/10800