Currently when we have a test we initially run a soft launch of 90% (original) / 10% (test) for a day, and if no issues occur we full launch the test to 50%/50%. Between Soft to Full launch we would have to clone the soft launch test to avoid the visitor bucketing issue.
It would be nice to see this implemented, and not have many archived tests that are similar - e.g. dev version, soft launch version, full version, then we will have the Winner version until the developers implement it directly to the site.
Maybe a "Reset" visitors option, everytime we adjust the traffic allocation?