Now that the test has been restarted, we should verify that the data is coming in correctly and as expected.
Sampling within both wikis seems fine: https://github.com/wikimedia-research/CompletionSuggestionTest/blob/master/data_validation/wiki.png
I don't like that the group proportions within dewiki are just barely within 2 standard deviations of each other. They should be a lot closer to 50/50 than 55/45.
P.S. @Ironholds we need to start using error bars for our proportions as in https://raw.githubusercontent.com/wikimedia-research/CompletionSuggestionTest/master/data_validation/wiki.png