When an A/B test distribution is changed to 100% for one variant, all customers see that variant regardless of which group they were originally assigned to. However, the system does not record an A/B test attribution event for customers during this 100% period - their variant assignment is not persisted to their profile.
What happens after the distribution is changed again
If the distribution is later changed back (for example, to 50/50), the system assigns these customers to a variant again as if for the first time.
Since no attribution was saved while the 100% variant was showing, a customer may now be placed in a different group than the one they were in before the 100% change.
Having now been exposed to more than one variant, these customers fall into the "Mixed" group and should be excluded from the A/B test evaluation.