So far the test utilities we built for the suggestions were made as we go following strict requirements of the tests being written. As we added more, they got bigger and with weirder APIs, sometimes duplicating code.

For example we have multiple ways to create dummy suggestions. We mix mocks and fakes for uiDelegate and SuggestionsSource, relying on Mockito or utility methods to set them up in similar ways.

We probably could clean it up by going from the interface specification, implementing an alternative to the native code properly and exposing straightforward ways to control their internal data.