Don't you think that this is a calculated trade-off? You can roll out the A/B testing to 2%, or 10%, or 50%. There are statistics on all this. Individual user feedback is irrelevant.