I am currently working on planning a test strategy for test automation for our mobile apps. The problem is that all the automation frameworks I've seen utilize a slightly different context than what the application really uses.
For example, I am evaluating SeeTest. To run automation tests on an non-jailbreak device the developer needs to add some frameworks to the project and actually build a different app the release app. So, when testing an instrumented iOS app with SeeTest, the app under test is not the same as the released app, which means that a full regression should be made on the released build.
This is the same of every framework I have seen.
How did others solve this issues?