The current method employed at our company is to have a long-lived release branch that we all develop on, with developers checking in any non-ab test code into both the branch as well as the trunk, while ab test code only gets checked into the branch. Then, once we release, winning test cells get merged into the trunk, while the majority of ab test code gets thrown away. This has been working for us for some time now, but we're starting a new project and I'm wondering if we might be able to improve our workflow with this new project. Having everyone checking in to both a branch and the trunk is a pain. Ideally, we'd have the more standard approach of developing against the trunk and branching for stabilization, but this would mean polluting our trunk with ab test code.