This is absolutely spot-on. Just last week, I have experienced at least 3 major issues because of "experimental" features getting pushed haphasardly. A much more complex issue is that A/B testing frequently causes feature interaction and feature layering bugs, because implementing a feature in a way that is consistent with the rest of the system goes contrary to both the short-term incentives of the PM/team pushing the feature ("let's throw it at the wall quickly") and the spirit of A/B testing ("rapid experimentation"). It is quite sad, really.