A good A/B test is often "sticky" (usually using cookies); once you're in a certain bucket, you stay in that bucket. So a different browser could conceivably make a difference.