This used to be normal, you'd see an email come out a couple weeks in advance and a banner notice on the site, and then a site would go down on like a Sunday or Monday morning night between 1 and 3-4am ET in the US to try and minimize disruption.
The key reasons were as others said, but in my personal experience generally data service/schema related where a key data source would be offline for a period of time was the primary cause. One thing not mentioned is hardware upgrades has always been a key driver of this type of maintenance. Self hosted systems have to be transferred across hardware and brought back online and tested. This still happens today for many larger enterprises as upgrading to the latest hardware to support your SQL Server or Oracle database comes down to cost. They can go down for 2-3 hours for the switchover, OR spend 6-8 weeks engineering & testing a non-downtime failover. And the math is simple, most businesses don't have a true 24/7 need to be online and the don't make money 24/7 online, so this is still a solid reason and solution in those cases.
At one company I consulted at we gave them an option to upgrade their relational database solution (hardware + software) with 0 downtime, but it would take around 700 man hours to setup and test, or they could go offline for `4 hours and be back online. They were a B2B company and said the difference in cost was not worth it and went down for a few hours over a weekend instead, which honestly made far more sense to me. They did have clients that accessed the system overnight, but they just sent out a notice and no one cared.