Ease A-B Testing / Beta Testing support within a framework

As no one's answered this question, I'll attempt to do so.

Basically, I'm not sure if there's a directly useful connection between your PHP framework and your A/B testing needs. I think this is mainly because what you're testing can be almost anything: the colour of a conversion-sensitive button, a page layout, an entire registration funnel, etc. These don't inherently have anything to do with your PHP framework and there are lots of options for how you could do your testing.

Another issue is that you might not really know the parameters of what you're testing until you start testing. Your testing might lead you down a way that you didn't really even consider, so how could you have accounted for it in how you built the site? If you need a REALLY wide window for what you'll be testing, you're probably better off not building it at all and using some type of vapor/smoke-testing to get the basic concepts right first. Not everything can be subjected to testing and you'll still need subjectively-generated hypotheses as your test cases (and your testing will be only as good as your hypotheses).

If you have something very specific that you need to test repeatedly over time and want to build this flexibility into the system, then I'd look for the most obvious solution in the framework to make it happen. For example, if you're using Symfony and if you think that you'll need to test 50 different sidebar variations for a page over the course of 6 months, it probably makes sense to build it as a slot/component so you can build some logic around simplifying your testing and swap those sidebars with ease. I'm not sure why it would need to be anything more complicated than that.

Overall, I'd also add that the role of A/B testing should to guide your product to sell/convert/monetize/engage better. Unless you're building some type of a testing platform, I wouldn't over-think it. I tend to see that most sites fail to test sufficiently not because the system isn't flexible enough for various test cases but because top management won't give enough product/dev time for it, or because people aren't making enough use of their analytics packages to draw even the most basic of conclusions.

Hope that helps.