Definitely can repro.

Starting working on it.

Hum actually it might have been a fluke. I tried to repro several times and could not succeed. I think I might have triggered it from a syncing state when I tried to repro initially.

As for the metrics, I think you are filtering for users that have the flag on correct? Not that they are in the Butter state?

That means that you are also seeing the data when those users are in full sync? This would explain why we see wallet addresses added.

Thanks for investigating, Seb!

You are right! I am sorry for the confusion; the metrics seem okay.
(https://uma.googleplex.com/p/chrome/variations/?sid=92b4be59eba0818817a4e4e0d9abb171)

Even though there Autofill.WalletAddressesAdded has 150% increase in samples in the Enabled_Butter group compared to Control, this is consistent with the Enabled_WalletUSS group and simply corresponds to Directory -> USS conversion. These are users in full sync mode.

The reason to believe so is the comparison with the metric Autofill.WalletCardsAdded. Again, Enabled_WalletUSS has increased samples count (+180%) due to the same reason. Enabled_Butter really shows the impact of butter: +4000% increase in samples count.

Closing as it works as expected. Please re-open if you find counter-evidence.