The distribution toggle allows you control the portion of users randomly assigned to each bucket.
You can also control this by clicking on the percentage beneath the variable name and manually typing the percent amount.
The distribution can impact the time it takes for your test to reach statistical significance, especially if a very small portion of the audience is exposed to a variant being tested.
Note: It's generally best to avoid changing distribution mid-test. Doing so can threaten the integrity of your results, and it may cause your test to take longer to reach statistical significance.
If you do change the distribution of an active A/B test, it won't affect the users that have already been sorted into each variant (bucket). Only users who enter the test after the distribution change will be sorted according to the new distribution.