I can't remember where I read, listened or watched that the team that made the first successful plane did so by using a prototype that could easily be built, tested and improved (whereas other teams were going with long-and-costly-to-build planes).
We are using the same state of mind for our whatever-test-paradigm. We could try to make the perfect tests plan and implement it using the best iOS tests framework but one can't forget that there is no perfection out there.
Here the v1 of our testbed (1). Cardboard, adhesive tapes and $1 cases. A little hub inside, plugged into the MacMini you can see behind. The so-easy-to-setup Xcode Server, and here you are ! Cost: nearly $0.
But we had a few problems with it. Putting and removing the phones from their $1 cases was a bit of trouble for something you would like to do a few times a day (because this testbed isn't yet official, we are using the devices used for programming).
The hub wasn't easily accessible (and we all know accessibility is important), the testbed was too large and too fragile (the middle section began to cave in).
So here's the v2:
Total cost: less than $10, one hour of work. This is still a work-in-progress. The edges are rough and the other side has imperfection on alignment between top and side. BUT this is rock-solid, we put velcro between the wood and the cases (thus the phones can stay in their cases all day long) and the hub is directly accessible.
This object, not beautiful but not ugly, is in the middle of our open-space and reminds everyone that continuous integration is a real thing and that we have to improve its usage in our workflow, iteration after iteration.
(1): to be honest, this is the v2. But the real v1 is so ugly, it wouldn't have been decent to show it to you.