So I have refactored #16 slightly and also added a test suite for this bug:

• creating new resource that is not a fork, i.e. ensure we don't break original behaviour: TestResourceCreate.test_not_fork_resource_create()
• creating new resource that is fork: already done by pre-existing tests
• changing fork resource to non-forked resource, i.e. ensure we don't break the ability to override a fork resource with non-fork data: TestResourceUpdate.test_fork_resource_update_with_new_non_fork_resource_details()
• changing non-fork resource to fork: TestResourceUpdate.test_non_fork_resource_update_with_new_fork_details()
• syncing fork resource after parent change: this is essentially tested by TestResourceUpdate:test_fork_resource_update_with_activity_id()
• changing fields like description and permissions on forked resources: TestResourceUpdate: test_fork_resource_update_with_new_metadata()
• uploading new file to a fork resource, overriding the forkiness but using the API instead of UI, i.e. the main bug we were trying to fix: TestResourceUpdate.test_non_fork_resource_update_with_new_fork_details_via_api_call()

Also:

• renamed the test classes as the really calls to resource_update etc., which under the hood calls package_update etc. - this could be a preference thing but it just felt weird to have classes that said they were testing package actions but that never called package actions directly
• fixed a typo on (what is now) line 308
• added fork to four pre-existing test names for consistency with new tests