At the moment, if you change the split ratio after an experiment ran for a while and assign 0% to one of the variants, you'll still get some traffic from users that have potentially persisted that variant in their local storage.
Expected Behaviour
If we pull a variant from an experiment, it should not get any traffic anymore.
Actual Behaviour
The variant still gets traffic once the choice was persisted to the local storage.
At the moment, if you change the split ratio after an experiment ran for a while and assign 0% to one of the variants, you'll still get some traffic from users that have potentially persisted that variant in their local storage.
Expected Behaviour
If we pull a variant from an experiment, it should not get any traffic anymore.
Actual Behaviour
The variant still gets traffic once the choice was persisted to the local storage.