It wasn't abrupt, tzdata 2018c from January already had that change: https://access.redhat.com/errata/RHEA-2018:0232 https://access.redhat.com/errata/RHEA-2018:

It wasn't abrupt, tzdata 2018c from January already had that change: https://access.redhat.com/errata/RHEA-2018:0232 https://access.redhat.com/errata/RHEA-2018:0232 What was abrupt was that this month they decided to delay it even more due to the ENEM (https://pt.wikipedia.org/wiki/Exame_Nacional_do_Ensino_M%C3%A9dio https://pt.wikipedia.org/wiki/Exame_Nacional_do_Ensino_M%C3%...), and a few days later went back on that, keeping only the previous delay due to the elections. What happened with smartphones is that most are configured to automatically track the timezone from the carrier, and some carriers didn't set it up correctly. So when the normal date for DST arrived this Sunday, these phones entered daylight savings time, even though this should happen only two weeks later. (And of course, there's always the servers with unpatched software from 2016, and the servers with the Oracle JVM which for some reason uses its own copy of the tzdata database...)