If they had all those metrics and that many test users, how could such obvious backtick issues exist? Even simple dog-fooding by their own internal coding tea

If they had all those metrics and that many test users, how could such obvious backtick issues exist? Even simple dog-fooding by their own internal coding team should have revealed that issue. If they had a test group, it was definitely a group that didn't use those features to begin with and basically didn't use them in the new UI either. Meticulously tracking the wrong sample group simply leads to reams of bad data.