While we had a few end to end browser tests they tended to be slow so the pattern we ended up at was to generate output from the API to use as mock data in the UI tests.
I think it usually make sense to test the business logic at the JSON API level (except for particularly complicated parts where unit testing of that complex logic may make sense) and then use snapshot tests for the display logic.
I think it usually make sense to test the business logic at the JSON API level (except for particularly complicated parts where unit testing of that complex logic may make sense) and then use snapshot tests for the display logic.